2024-10-24 15:00:38 +00:00
<!doctype html>
< html lang = "zh" class = "no-js" >
< head >
< meta charset = "utf-8" >
< meta name = "viewport" content = "width=device-width,initial-scale=1" >
< meta name = "description" content = "Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)" >
< meta name = "author" content = "PaddleOCR PMC" >
< link rel = "canonical" href = "https://paddlepaddle.github.io/PaddleOCR/v2.9/FAQ.html" >
< link rel = "prev" href = "datasets/kie_datasets.html" >
< link rel = "next" href = "community/community_contribution.html" >
< link rel = "icon" href = "static/images/logo.jpg" >
2024-12-12 07:39:44 +00:00
< meta name = "generator" content = "mkdocs-1.6.1, mkdocs-material-9.5.48" >
2024-10-24 15:00:38 +00:00
< title > FAQ - PaddleOCR 文档< / title >
2024-12-12 07:39:44 +00:00
< link rel = "stylesheet" href = "assets/stylesheets/main.6f8fc17f.min.css" >
2024-10-24 15:00:38 +00:00
< link rel = "stylesheet" href = "assets/stylesheets/palette.06af60db.min.css" >
< style > : r o o t { - - m d - a d m o n i t i o n - i c o n - - n o t e : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 1 7 . 7 7 5 V 2 . 7 5 C 1 1 . 7 8 4 1 . 7 8 4 1 2 . 7 5 1 h 5 . 0 2 5 c . 4 6 4 0 . 9 1 . 1 8 4 1 . 2 3 8 . 5 1 3 l 6 . 2 5 6 . 2 5 a 1 . 7 5 1 . 7 5 0 0 1 0 2 . 4 7 4 l - 5 . 0 2 6 5 . 0 2 6 a 1 . 7 5 1 . 7 5 0 0 1 - 2 . 4 7 4 0 l - 6 . 2 5 - 6 . 2 5 A 1 . 7 5 1 . 7 5 0 0 1 1 7 . 7 7 5 m 1 . 5 0 c 0 . 0 6 6 . 0 2 6 . 1 3 . 0 7 3 . 1 7 7 l 6 . 2 5 6 . 2 5 a . 2 5 . 2 5 0 0 0 . 3 5 4 0 l 5 . 0 2 5 - 5 . 0 2 5 a . 2 5 . 2 5 0 0 0 0 - . 3 5 4 l - 6 . 2 5 - 6 . 2 5 a . 2 5 . 2 5 0 0 0 - . 1 7 7 - . 0 7 3 H 2 . 7 5 a . 2 5 . 2 5 0 0 0 - . 2 5 . 2 5 Z M 6 5 a 1 1 0 1 1 0 2 1 1 0 0 1 0 - 2 " / > < / s v g > ' ) ; - - m d - a d m o n i t i o n - i c o n - - a b s t r a c t : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 2 . 5 1 . 7 5 v 1 1 . 5 c 0 . 1 3 8 . 1 1 2 . 2 5 . 2 5 . 2 5 h 3 . 1 7 a . 7 5 . 7 5 0 0 1 0 1 . 5 H 2 . 7 5 A 1 . 7 5 1 . 7 5 0 0 1 1 1 3 . 2 5 V 1 . 7 5 C 1 . 7 8 4 1 . 7 8 4 0 2 . 7 5 0 h 8 . 5 C 1 2 . 2 1 6 0 1 3 . 7 8 4 1 3 1 . 7 5 v 7 . 7 3 6 a . 7 5 . 7 5 0 0 1 - 1 . 5 0 V 1 . 7 5 a . 2 5 . 2 5 0 0 0 - . 2 5 - . 2 5 h - 8 . 5 a . 2 5 . 2 5 0 0 0 - . 2 5 . 2 5 m 1 3 . 2 7 4 9 . 5 3 7 z l - 4 . 5 5 7 4 . 4 5 a . 7 5 . 7 5 0 0 1 - 1 . 0 5 5 - . 0 0 8 l - 1 . 9 4 3 - 1 . 9 5 a . 7 5 . 7 5 0 0 1 1 . 0 6 2 - 1 . 0 5 8 l 1 . 4 1 9 1 . 4 2 5 4 . 0 2 6 - 3 . 9 3 2 a . 7 5 . 7 5 0 1 1 1 . 0 4 8 1 . 0 7 4 M 4 . 7 5 4 h 4 . 5 a . 7 5 . 7 5 0 0 1 0 1 . 5 h - 4 . 5 a . 7 5 . 7 5 0 0 1 0 - 1 . 5 M 4 7 . 7 5 A . 7 5 . 7 5 0 0 1 4 . 7 5 7 h 2 a . 7 5 . 7 5 0 0 1 0 1 . 5 h - 2 A . 7 5 . 7 5 0 0 1 4 7 . 7 5 " / > < / s v g > ' ) ; - - m d - a d m o n i t i o n - i c o n - - i n f o : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 0 8 a 8 8 0 1 1 1 6 0 A 8 8 0 0 1 0 8 m 8 - 6 . 5 a 6 . 5 6 . 5 0 1 0 0 1 3 6 . 5 6 . 5 0 0 0 0 - 1 3 M 6 . 5 7 . 7 5 A . 7 5 . 7 5 0 0 1 7 . 2 5 7 h 1 a . 7 5 . 7 5 0 0 1 . 7 5 . 7 5 v 2 . 7 5 h . 2 5 a . 7 5 . 7 5 0 0 1 0 1 . 5 h - 2 a . 7 5 . 7 5 0 0 1 0 - 1 . 5 h . 2 5 v - 2 h - . 2 5 a . 7 5 . 7 5 0 0 1 - . 7 5 - . 7 5 M 8 6 a 1 1 0 1 1 0 - 2 1 1 0 0 1 0 2 " / > < / s v g > ' ) ; - - m d - a d m o n i t i o n - i c o n - - t i p : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 3 . 4 9 9 . 7 5 a . 7 5 . 7 5 0 0 1 1 . 5 0 v . 9 9 6 C 5 . 9 2 . 9 0 3 6 . 7 9 3 3 . 6 5 7 . 6 6 2 4 . 3 7 6 l . 2 4 . 2 0 2 c - . 0 3 6 - . 6 9 4 . 0 5 5 - 1 . 4 2 2 . 4 2 6 - 2 . 1 6 3 C 9 . 1 . 8 7 3 1 0 . 7 9 4 - . 0 4 5 1 2 . 6 2 2 . 2 6 1 4 . 4 0 8 . 5 5 8 1 6 1 . 9 4 1 6 4 . 2 5 c 0 1 . 2 7 8 - . 9 5 4 2 . 5 7 5 - 2 . 4 4 2 . 7 3 4 l . 1 4 6 . 5 0 8 . 0 6 5 . 2 2 c . 2 0 3 . 7 0 1 . 4 1 2 1 . 4 5 5 . 4 7 6 2 . 2 2 6 . 1 4 2 1 . 7 0 7 - . 4 3 . 0 3 - 1 . 4 8 7 3 . 8 9 8 C 1 1 . 7 1 4 1 4 . 6 7 1 1 0 . 2 7 1 5 8 . 7 5 1 5 h - 6 a . 7 5 . 7 5 0 0 1 0 - 1 . 5 h 1 . 3 7 6 a 4 . 5 4 . 5 0 0 1 - . 5 6 3 - 1 . 1 9 1 3 . 8 4 3 . 8 4 0 0 1 - . 0 5 - 2 . 0 6 3 4 . 6 5 4 . 6 5 0 0 1 - 2 . 0 2 5 - . 2 9 3 . 7 5 . 7 5 0 0 1 . 5 2 5 - 1 . 4 0 6 c 1 . 3 5 7 . 5 0 7 2 . 3 7 6 - . 0 0 6 2 . 6 9 8 - . 3 1 8 l . 0 0 9 - . 0 1 a . 7 4 7 . 7 4 7 0 0 1 1 . 0 6 0 . 7 5 . 7 5 0 0 1 - . 0 1 2 1 . 0 7 4 c - . 9 1 2 . 9 2 - . 9 9 2 1 . 8 3 5 - . 7 6 8 2 . 5 8 6 . 2 2 1 . 7 4 . 7 4 5 1 . 3 3 7 1 . 1 9 6 1 . 6 2 1 H 8 . 7 5 c 1 . 3 4 3 0 2 . 3 9 8 - . 2 9 6 3 . 0 7 4 - . 8 3 6 . 6 3 5 - . 5 0 7 1 . 0 3 6 - 1 . 3 1 . 9 2 8 - 2 . 6 0 2 - . 0 5 - . 6 0 3 - . 2 1 6 - 1 . 2 2 4 - . 4 2 2 - 1 . 9 3 l - . 0 6 4 - . 2 2 1 c - . 1 2 - . 4 0 7 - . 2 4 6 - . 8 4 - . 3 5 3 - 1 . 2 9 a 2 . 4 2 . 4 0 0 1 - . 5 0 7 - . 4 4 1 3 . 1 3 . 1 0 0 1 - . 6 3 3 - 1 . 2 4 8 . 7 5 . 7 5 0 0 1 1 . 4 5 5 - . 3 6 4 c . 0 4 6 . 1 8 5 . 1 4 4 . 4 3 6 . 3 1 . 6 2 7 . 1 4 6 . 1 6 8 . 3 5 3 . 3 0 5 . 7 1 2 . 3 0 5 . 7 3 8 0 1 . 2 5 - . 6 1 5 1 . 2 5 - 1 . 2 5 0 - 1 . 4 7 - . 9 5 - 2 . 3 1 5 - 2 . 1 2 3 - 2 . 5 1 - 1 . 1 7 2 - . 1 9 6 - 2 . 2 2 7 . 3 8 7 - 2 . 7 0 6 1 . 3 4 5 - . 4 6 . 9 2 - . 2 7 1 . 7 7 4 . 0 1 9 3 . 0 6 2 l . 0 4 2 . 1 9 . 0 1 . 0 5 c . 3 4 8 . 4 4 3 . 6 6 6 . 9 4 9 . 9 4 1 . 5 5 3 a . 7 5 . 7 5 0 1 1 - 1 . 3 6 5 . 6 2 c - . 5 5 3 - 1 . 2 1 7 - 1 . 3 2 - 1 . 9 4 - 2 . 3 - 2 . 7 6 8 L 6 . 7 5 . 5 2 7 c - . 8 1 4 - . 6 8 - 1 . 7 5 - 1 . 4 6 2 - 2 . 6 9 2 - 2 . 6 1 9 a 3 . 7 3 . 7 0 0 0 - 1 . 0 2 3 . 8 8 c - . 4 0 6 . 4 9 5 - . 6 6 3 1 . 0 3 6 - . 7 2 2 1 . 5 0 8 . 1 1 6 . 1 2 2 . 3 0 6 . 2 1 . 5 9 1 . 2 3 9 . 3 8 8 . 0 3 8 . 7 9 7 - . 0 6 1 . 0 3 2 - . 1 9 a . 7 5 . 7 5 0 0 1 . 7 2 8 1 . 3 1 c - . 5 1 5 . 2 8 7 - 1 . 2 3 . 4 3 9 - 1 . 9 0 6 . 3 7 3 - . 6 8 2 - . 0 6 7 - 1 . 4 7 3 - . 3 8 - 1 . 8 7 9 - 1 . 1 9 3 L . 7 5 5 . 6 7 7 V 5 . 5 c 0 - . 9 8 4 . 4 8 - 1 . 9 4 1 . 0 7 7 - 2 . 6 6 4 . 4 6 - . 5 5 9 1 . 0 5 - 1 . 0 5 5 1 . 6 7 3 - 1 . 3 5 3 z " / > < / s v g > ' ) ; - - m d - a d m o n i t i o n - i c o n - - s u c c e s s : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 1 3 . 7 8 4 . 2 2 a . 7 5 . 7 5 0 0 1 0 1 . 0 6 l - 7 . 2 5 7 . 2 5 a . 7 5 . 7 5 0 0 1 - 1 . 0 6 0 L 2 . 2 2 9 . 2 8 a . 7 5 . 7 5 0 0 1 . 0 1 8 - 1 . 0 4 2 . 7 5 . 7 5 0 0 1 1 . 0 4 2 - . 0 1 8 L 6 1 0 . 9 4 l 6 . 7 2 - 6 . 7 2 a . 7 5 . 7 5 0 0 1 1 . 0 6 0 " / > < / s v g > ' ) ; - - m d - a d m o n i t i o n - i c o n - - q u e s t i o n : u r l ( ' d a t a : i m a g e / s v g + x m l ; c h a r s e t = u t f - 8 , < s v g x m l n s = " h t t p : / / w w w . w 3 . o r g / 2 0 0 0 / s v g " v i e w B o x = " 0 0 1 6 1 6 " > < p a t h d = " M 0 8 a 8 8 0 1 1 1 6 0 A 8 8 0 0 1 0 8 m 8 - 6 . 5 a 6 . 5 6 . 5 0 1 0 0 1 3 6 . 5 6 . 5 0 0 0 0 - 1 3 M 6 . 9 2 6 . 0 8 5 h . 0 0 1 a . 7 4 9 . 7 4 9 0 1 1 - 1 . 3 4 2 - . 6 7 c . 1 6 9 - . 3 3 9 . 4 3 6 - . 7 0 1 . 8 4 9 - . 9 7 7 C 6 . 8 4 5 4 . 1 6 7 . 3 6 9 4 8 4 a 2 . 7 6 2 . 7 6 0 0 1 1 . 6 3 7 . 5 2 5 c . 5 0 3 . 3 7 7 . 8 6 3 . 9 6 5 . 8 6 3 1 . 7 2 5 0 . 4 4 8 - . 1 1 5 . 8 3 - . 3 2 9 1 . 1 5 - . 2 0 5 . 3 0 7 - . 4 7 . 5 1 3 - . 6 9 2 . 6 6 2 - . 1 0 9 . 0 7 2 - . 2 2 . 1 3 8 - . 3 1 3 . 1 9 5 l - . 0 0 6 . 0 0 4 a 6 6 0 0 0 - . 2 6 . 1 6 1 1 0 0 0 - . 2 7 6 . 2 4 5 . 7 5 . 7 5 0 0 1 - 1 . 2 4 8 - . 8 3 2 c . 1 8 4 - . 2 6 4 . 4 2 - . 4 8 9 . 6 9 2 - . 6 6 1 q . 1 5 4 - . 1 . 3 1 3 - . 1 9 5 l . 0 0 7 - . 0 0 4 c . 1 - . 0 6 1 . 1 8 2 - . 1 1 . 2 5 8 - . 1 6 1 a 1 1 0 0 0 . 2 7 7 - . 2 4 5 C 8 . 9 6 6 . 5 1 4 9 6 . 4 2 7 9 6 . 2 5 a . 6 1 . 6 1 0 0 0 - . 2 6 2 - . 5 2 5 A 1 . 2 7 1 . 2 7 0 0 0 8 5 . 5 c - . 3 6 9 0 - . 5 9 5 . 0 9 - . 7 4 . 1 8 7 a 1 1 0 0 0 - . 3 4 . 3 9 8 M 9 1 1 a 1 1 0 1 1 - 2 0 1 1 0 0 1 2 0 " / > < / s v g > ' ) ; - - m d -
2024-12-12 07:39:44 +00:00
< style > . md-tag . md-tag--default-tag { --md-tag-icon : url ( 'data:image/svg+xml;charset=utf-8,<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 512"><!--! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc.--><path d="M0 80v149.5c0 17 6.7 33.3 18.7 45.3l176 176c25 25 65.5 25 90.5 0l133.5-133.5c25-25 25-65.5 0-90.5l-176-176c-12-12-28.3-18.7-45.3-18.7L48 32C21.5 32 0 53.5 0 80m112 32a32 32 0 1 1 0 64 32 32 0 1 1 0-64"/></svg>' ) ; } . md-tag . md-tag--hardware-tag { --md-tag-icon : url ( 'data:image/svg+xml;charset=utf-8,<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 512 512"><!--! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc.--><path d="M176 24c0-13.3-10.7-24-24-24s-24 10.7-24 24v40c-35.3 0-64 28.7-64 64H24c-13.3 0-24 10.7-24 24s10.7 24 24 24h40v56H24c-13.3 0-24 10.7-24 24s10.7 24 24 24h40v56H24c-13.3 0-24 10.7-24 24s10.7 24 24 24h40c0 35.3 28.7 64 64 64v40c0 13.3 10.7 24 24 24s24-10.7 24-24v-40h56v40c0 13.3 10.7 24 24 24s24-10.7 24-24v-40h56v40c0 13.3 10.7 24 24 24s24-10.7 24-24v-40c35.3 0 64-28.7 64-64h40c13.3 0 24-10.7 24-24s-10.7-24-24-24h-40v-56h40c13.3 0 24-10.7 24-24s-10.7-24-24-24h-40v-56h40c13.3 0 24-10.7 24-24s-10.7-24-24-24h-40c0-35.3-28.7-64-64-64V24c0-13.3-10.7-24-24-24s-24 10.7-24 24v40h-56V24c0-13.3-10.7-24-24-24s-24 10.7-24 24v40h-56zm-16 104h192c17.7 0 32 14.3 32 32v192c0 17.7-14.3 32-32 32H160c-17.7 0-32-14.3-32-32V160c0-17.7 14.3-32 32-32m192 32H160v192h192z"/></svg>' ) ; } . md-tag . md-tag--software-tag { --md-tag-icon : url ( 'data:image/svg+xml;charset=utf-8,<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 640 512"><!--! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc.--><path d="M64 96c0-35.3 28.7-64 64-64h384c35.3 0 64 28.7 64 64v256h-64V96H128v256H64zM0 403.2C0 392.6 8.6 384 19.2 384h601.6c10.6 0 19.2 8.6 19.2 19.2 0 42.4-34.4 76.8-76.8 76.8H76.8C34.4 480 0 445.6 0 403.2M281 209l-31 31 31 31c9.4 9.4 9.4 24.6 0 33.9s-24.6 9.4-33.9 0l-48-48c-9.4-9.4-9.4-24.6 0-33.9l48-48c9.4-9.4 24.6-9.4 33.9 0s9.4 24.6 0 33.9zm112-34 48 48c9.4 9.4 9.4 24.6 0 33.9l-48 48c-9.4 9.4-24.6 9.4-33.9 0s-9.4-24.6 0-33.9l31-31-31-31c-9.4-9.4-9.4-24.6 0-33.9s24.6-9.4 33.9 0z"/></svg>' ) ; } < / style >
2024-10-24 15:00:38 +00:00
< script src = "https://unpkg.com/iframe-worker/shim" > < / script >
< link rel = "preconnect" href = "https://fonts.gstatic.com" crossorigin >
< link rel = "stylesheet" href = "https://fonts.googleapis.com/css?family=Roboto:300,300i,400,400i,700,700i%7CRoboto+Mono:400,400i,700,700i&display=fallback" >
< style > : root { --md-text-font : "Roboto" ; --md-code-font : "Roboto Mono" } < / style >
< link rel = "stylesheet" href = "https://unpkg.com/katex@0/dist/katex.min.css" >
< script > _ _md _scope = new URL ( "." , location ) , _ _md _hash = e => [ ... e ] . reduce ( ( ( e , _ ) => ( e << 5 ) - e + _ . charCodeAt ( 0 ) ) , 0 ) , _ _md _get = ( e , _ = localStorage , t = _ _md _scope ) => JSON . parse ( _ . getItem ( t . pathname + "." + e ) ) , _ _md _set = ( e , _ , t = localStorage , a = _ _md _scope ) => { try { t . setItem ( a . pathname + "." + e , JSON . stringify ( _ ) ) } catch ( e ) { } } < / script >
< / head >
< body dir = "ltr" data-md-color-scheme = "default" data-md-color-primary = "indigo" data-md-color-accent = "indigo" >
< input class = "md-toggle" data-md-toggle = "drawer" type = "checkbox" id = "__drawer" autocomplete = "off" >
< input class = "md-toggle" data-md-toggle = "search" type = "checkbox" id = "__search" autocomplete = "off" >
< label class = "md-overlay" for = "__drawer" > < / label >
< div data-md-component = "skip" >
< a href = "#1" class = "md-skip" >
跳转至
< / a >
< / div >
< div data-md-component = "announce" >
< / div >
< div data-md-color-scheme = "default" data-md-component = "outdated" hidden >
< aside class = "md-banner md-banner--warning" >
< div class = "md-banner__inner md-grid md-typeset" >
You're not viewing the latest version.
< a href = "../." >
< strong > Click here to go to latest.< / strong >
< / a >
< / div >
< script > var el = document . querySelector ( "[data-md-component=outdated]" ) , outdated = _ _md _get ( "__outdated" , sessionStorage ) ; ! 0 === outdated && el && ( el . hidden = ! 1 ) < / script >
< / aside >
< / div >
< header class = "md-header md-header--shadow md-header--lifted" data-md-component = "header" >
< nav class = "md-header__inner md-grid" aria-label = "页眉" >
< a href = "index.html" title = "PaddleOCR 文档" class = "md-header__button md-logo" aria-label = "PaddleOCR 文档" data-md-component = "logo" >
< img src = "static/images/logo.jpg" alt = "logo" >
< / a >
< label class = "md-header__button md-icon" for = "__drawer" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M3 6h18v2H3zm0 5h18v2H3zm0 5h18v2H3z" / > < / svg >
< / label >
< div class = "md-header__title" data-md-component = "header-title" >
< div class = "md-header__ellipsis" >
< div class = "md-header__topic" >
< span class = "md-ellipsis" >
PaddleOCR 文档
< / span >
< / div >
< div class = "md-header__topic" data-md-component = "header-topic" >
< span class = "md-ellipsis" >
FAQ
< / span >
< / div >
< / div >
< / div >
< form class = "md-header__option" data-md-component = "palette" >
< input class = "md-option" data-md-color-media = "(prefers-color-scheme: light)" data-md-color-scheme = "default" data-md-color-primary = "indigo" data-md-color-accent = "indigo" aria-label = "Switch to dark mode" type = "radio" name = "__palette" id = "__palette_0" >
< label class = "md-header__button md-icon" title = "Switch to dark mode" for = "__palette_1" hidden >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M12 8a4 4 0 0 0-4 4 4 4 0 0 0 4 4 4 4 0 0 0 4-4 4 4 0 0 0-4-4m0 10a6 6 0 0 1-6-6 6 6 0 0 1 6-6 6 6 0 0 1 6 6 6 6 0 0 1-6 6m8-9.31V4h-4.69L12 .69 8.69 4H4v4.69L.69 12 4 15.31V20h4.69L12 23.31 15.31 20H20v-4.69L23.31 12z" / > < / svg >
< / label >
< input class = "md-option" data-md-color-media = "(prefers-color-scheme: dark)" data-md-color-scheme = "slate" data-md-color-primary = "black" data-md-color-accent = "indigo" aria-label = "Switch to system preference" type = "radio" name = "__palette" id = "__palette_1" >
< label class = "md-header__button md-icon" title = "Switch to system preference" for = "__palette_0" hidden >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M12 18c-.89 0-1.74-.2-2.5-.55C11.56 16.5 13 14.42 13 12s-1.44-4.5-3.5-5.45C10.26 6.2 11.11 6 12 6a6 6 0 0 1 6 6 6 6 0 0 1-6 6m8-9.31V4h-4.69L12 .69 8.69 4H4v4.69L.69 12 4 15.31V20h4.69L12 23.31 15.31 20H20v-4.69L23.31 12z" / > < / svg >
< / label >
< / form >
< script > var palette = _ _md _get ( "__palette" ) ; if ( palette && palette . color ) { if ( "(prefers-color-scheme)" === palette . color . media ) { var media = matchMedia ( "(prefers-color-scheme: light)" ) , input = document . querySelector ( media . matches ? "[data-md-color-media='(prefers-color-scheme: light)']" : "[data-md-color-media='(prefers-color-scheme: dark)']" ) ; palette . color . media = input . getAttribute ( "data-md-color-media" ) , palette . color . scheme = input . getAttribute ( "data-md-color-scheme" ) , palette . color . primary = input . getAttribute ( "data-md-color-primary" ) , palette . color . accent = input . getAttribute ( "data-md-color-accent" ) } for ( var [ key , value ] of Object . entries ( palette . color ) ) document . body . setAttribute ( "data-md-color-" + key , value ) } < / script >
< div class = "md-header__option" >
< div class = "md-select" >
< button class = "md-header__button md-icon" aria-label = "选择当前语言" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "m12.87 15.07-2.54-2.51.03-.03A17.5 17.5 0 0 0 14.07 6H17V4h-7V2H8v2H1v2h11.17C11.5 7.92 10.44 9.75 9 11.35 8.07 10.32 7.3 9.19 6.69 8h-2c.73 1.63 1.73 3.17 2.98 4.56l-5.09 5.02L4 19l5-5 3.11 3.11zM18.5 10h-2L12 22h2l1.12-3h4.75L21 22h2zm-2.62 7 1.62-4.33L19.12 17z" / > < / svg >
< / button >
< div class = "md-select__inner" >
< ul class = "md-select__list" >
< li class = "md-select__item" >
< a href = "FAQ.html" hreflang = "zh" class = "md-select__link" >
简体中文
< / a >
< / li >
< li class = "md-select__item" >
< a href = "en/FAQ.html" hreflang = "en" class = "md-select__link" >
English
< / a >
< / li >
< li class = "md-select__item" >
< a href = "https://github.com/PaddlePaddle/PaddleOCR/discussions/13374" hreflang = "null" class = "md-select__link" >
Help translating
< / a >
< / li >
< / ul >
< / div >
< / div >
< / div >
< label class = "md-header__button md-icon" for = "__search" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5" / > < / svg >
< / label >
< div class = "md-search" data-md-component = "search" role = "dialog" >
< label class = "md-search__overlay" for = "__search" > < / label >
< div class = "md-search__inner" role = "search" >
< form class = "md-search__form" name = "search" >
< input type = "text" class = "md-search__input" name = "query" aria-label = "搜索" placeholder = "搜索" autocapitalize = "off" autocorrect = "off" autocomplete = "off" spellcheck = "false" data-md-component = "search-query" required >
< label class = "md-search__icon md-icon" for = "__search" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M9.5 3A6.5 6.5 0 0 1 16 9.5c0 1.61-.59 3.09-1.56 4.23l.27.27h.79l5 5-1.5 1.5-5-5v-.79l-.27-.27A6.52 6.52 0 0 1 9.5 16 6.5 6.5 0 0 1 3 9.5 6.5 6.5 0 0 1 9.5 3m0 2C7 5 5 7 5 9.5S7 14 9.5 14 14 12 14 9.5 12 5 9.5 5" / > < / svg >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 320 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M41.4 233.4c-12.5 12.5-12.5 32.8 0 45.3l160 160c12.5 12.5 32.8 12.5 45.3 0s12.5-32.8 0-45.3L109.3 256l137.3-137.4c12.5-12.5 12.5-32.8 0-45.3s-32.8-12.5-45.3 0l-160 160z" / > < / svg >
2024-10-24 15:00:38 +00:00
< / label >
< nav class = "md-search__options" aria-label = "查找" >
< a href = "javascript:void(0)" class = "md-search__icon md-icon" title = "分享" aria-label = "分享" data-clipboard data-clipboard-text = "" data-md-component = "search-share" tabindex = "-1" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M18 16.08c-.76 0-1.44.3-1.96.77L8.91 12.7c.05-.23.09-.46.09-.7s-.04-.47-.09-.7l7.05-4.11c.54.5 1.25.81 2.04.81a3 3 0 0 0 3-3 3 3 0 0 0-3-3 3 3 0 0 0-3 3c0 .24.04.47.09.7L8.04 9.81C7.5 9.31 6.79 9 6 9a3 3 0 0 0-3 3 3 3 0 0 0 3 3c.79 0 1.5-.31 2.04-.81l7.12 4.15c-.05.21-.08.43-.08.66 0 1.61 1.31 2.91 2.92 2.91s2.92-1.3 2.92-2.91A2.92 2.92 0 0 0 18 16.08" / > < / svg >
< / a >
< button type = "reset" class = "md-search__icon md-icon" title = "清空当前内容" aria-label = "清空当前内容" tabindex = "-1" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M19 6.41 17.59 5 12 10.59 6.41 5 5 6.41 10.59 12 5 17.59 6.41 19 12 13.41 17.59 19 19 17.59 13.41 12z" / > < / svg >
< / button >
< / nav >
< div class = "md-search__suggest" data-md-component = "search-suggest" > < / div >
< / form >
< div class = "md-search__output" >
< div class = "md-search__scrollwrap" tabindex = "0" data-md-scrollfix >
< div class = "md-search-result" data-md-component = "search-result" >
< div class = "md-search-result__meta" >
正在初始化搜索引擎
< / div >
< ol class = "md-search-result__list" role = "presentation" > < / ol >
< / div >
< / div >
< / div >
< / div >
< / div >
< div class = "md-header__source" >
< a href = "https://github.com/PaddlePaddle/PaddleOCR" title = "前往仓库" class = "md-source" data-md-component = "source" >
< div class = "md-source__icon md-icon" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 496 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M165.9 397.4c0 2-2.3 3.6-5.2 3.6-3.3.3-5.6-1.3-5.6-3.6 0-2 2.3-3.6 5.2-3.6 3-.3 5.6 1.3 5.6 3.6m-31.1-4.5c-.7 2 1.3 4.3 4.3 4.9 2.6 1 5.6 0 6.2-2s-1.3-4.3-4.3-5.2c-2.6-.7-5.5.3-6.2 2.3m44.2-1.7c-2.9.7-4.9 2.6-4.6 4.9.3 2 2.9 3.3 5.9 2.6 2.9-.7 4.9-2.6 4.6-4.6-.3-1.9-3-3.2-5.9-2.9M244.8 8C106.1 8 0 113.3 0 252c0 110.9 69.8 205.8 169.5 239.2 12.8 2.3 17.3-5.6 17.3-12.1 0-6.2-.3-40.4-.3-61.4 0 0-70 15-84.7-29.8 0 0-11.4-29.1-27.8-36.6 0 0-22.9-15.7 1.6-15.4 0 0 24.9 2 38.6 25.8 21.9 38.6 58.6 27.5 72.9 20.9 2.3-16 8.8-27.1 16-33.7-55.9-6.2-112.3-14.3-112.3-110.5 0-27.5 7.6-41.3 23.6-58.9-2.6-6.5-11.1-33.3 2.6-67.9 20.9-6.5 69 27 69 27 20-5.6 41.5-8.5 62.8-8.5s42.8 2.9 62.8 8.5c0 0 48.1-33.6 69-27 13.7 34.7 5.2 61.4 2.6 67.9 16 17.7 25.8 31.5 25.8 58.9 0 96.5-58.9 104.2-114.8 110.5 9.2 7.9 17 22.9 17 46.4 0 33.7-.3 75.4-.3 83.6 0 6.5 4.6 14.4 17.3 12.1C428.2 457.8 496 362.9 496 252 496 113.3 383.5 8 244.8 8M97.2 352.9c-1.3 1-1 3.3.7 5.2 1.6 1.6 3.9 2.3 5.2 1 1.3-1 1-3.3-.7-5.2-1.6-1.6-3.9-2.3-5.2-1m-10.8-8.1c-.7 1.3.3 2.9 2.3 3.9 1.6 1 3.6.7 4.3-.7.7-1.3-.3-2.9-2.3-3.9-2-.6-3.6-.3-4.3.7m32.4 35.6c-1.6 1.3-1 4.3 1.3 6.2 2.3 2.3 5.2 2.6 6.5 1 1.3-1.3.7-4.3-1.3-6.2-2.2-2.3-5.2-2.6-6.5-1m-11.4-14.7c-1.6 1-1.6 3.6 0 5.9s4.3 3.3 5.6 2.3c1.6-1.3 1.6-3.9 0-6.2-1.4-2.3-4-3.3-5.6-2" / > < / svg >
2024-10-24 15:00:38 +00:00
< / div >
< div class = "md-source__repository" >
PaddlePaddle/PaddleOCR
< / div >
< / a >
< / div >
< / nav >
< nav class = "md-tabs" aria-label = "标签" data-md-component = "tabs" >
< div class = "md-grid" >
< ul class = "md-tabs__list" >
< li class = "md-tabs__item" >
< a href = "index.html" class = "md-tabs__link" >
Home
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "quick_start.html" class = "md-tabs__link" >
快速开始
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "update.html" class = "md-tabs__link" >
近期更新
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "paddlex/overview.html" class = "md-tabs__link" >
低代码全流程开发
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "model/index.html" class = "md-tabs__link" >
模型
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "ppocr/overview.html" class = "md-tabs__link" >
PP-OCR 文本检测识别
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "ppstructure/overview.html" class = "md-tabs__link" >
PP-Structure文档分析
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "algorithm/overview.html" class = "md-tabs__link" >
前沿算法与模型
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "applications/overview.html" class = "md-tabs__link" >
场景应用
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "data_anno_synth/overview.html" class = "md-tabs__link" >
数据标注与合成
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "datasets/datasets.html" class = "md-tabs__link" >
数据集
< / a >
< / li >
< li class = "md-tabs__item md-tabs__item--active" >
< a href = "FAQ.html" class = "md-tabs__link" >
FAQ
< / a >
< / li >
< li class = "md-tabs__item" >
< a href = "community/community_contribution.html" class = "md-tabs__link" >
社区
< / a >
< / li >
< / ul >
< / div >
< / nav >
< / header >
< div class = "md-container" data-md-component = "container" >
< main class = "md-main" data-md-component = "main" >
< div class = "md-main__inner md-grid" >
< div class = "md-sidebar md-sidebar--primary" data-md-component = "sidebar" data-md-type = "navigation" hidden >
< div class = "md-sidebar__scrollwrap" >
< div class = "md-sidebar__inner" >
< nav class = "md-nav md-nav--primary md-nav--lifted" aria-label = "导航栏" data-md-level = "0" >
< label class = "md-nav__title" for = "__drawer" >
< a href = "index.html" title = "PaddleOCR 文档" class = "md-nav__button md-logo" aria-label = "PaddleOCR 文档" data-md-component = "logo" >
< img src = "static/images/logo.jpg" alt = "logo" >
< / a >
PaddleOCR 文档
< / label >
< div class = "md-nav__source" >
< a href = "https://github.com/PaddlePaddle/PaddleOCR" title = "前往仓库" class = "md-source" data-md-component = "source" >
< div class = "md-source__icon md-icon" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 496 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M165.9 397.4c0 2-2.3 3.6-5.2 3.6-3.3.3-5.6-1.3-5.6-3.6 0-2 2.3-3.6 5.2-3.6 3-.3 5.6 1.3 5.6 3.6m-31.1-4.5c-.7 2 1.3 4.3 4.3 4.9 2.6 1 5.6 0 6.2-2s-1.3-4.3-4.3-5.2c-2.6-.7-5.5.3-6.2 2.3m44.2-1.7c-2.9.7-4.9 2.6-4.6 4.9.3 2 2.9 3.3 5.9 2.6 2.9-.7 4.9-2.6 4.6-4.6-.3-1.9-3-3.2-5.9-2.9M244.8 8C106.1 8 0 113.3 0 252c0 110.9 69.8 205.8 169.5 239.2 12.8 2.3 17.3-5.6 17.3-12.1 0-6.2-.3-40.4-.3-61.4 0 0-70 15-84.7-29.8 0 0-11.4-29.1-27.8-36.6 0 0-22.9-15.7 1.6-15.4 0 0 24.9 2 38.6 25.8 21.9 38.6 58.6 27.5 72.9 20.9 2.3-16 8.8-27.1 16-33.7-55.9-6.2-112.3-14.3-112.3-110.5 0-27.5 7.6-41.3 23.6-58.9-2.6-6.5-11.1-33.3 2.6-67.9 20.9-6.5 69 27 69 27 20-5.6 41.5-8.5 62.8-8.5s42.8 2.9 62.8 8.5c0 0 48.1-33.6 69-27 13.7 34.7 5.2 61.4 2.6 67.9 16 17.7 25.8 31.5 25.8 58.9 0 96.5-58.9 104.2-114.8 110.5 9.2 7.9 17 22.9 17 46.4 0 33.7-.3 75.4-.3 83.6 0 6.5 4.6 14.4 17.3 12.1C428.2 457.8 496 362.9 496 252 496 113.3 383.5 8 244.8 8M97.2 352.9c-1.3 1-1 3.3.7 5.2 1.6 1.6 3.9 2.3 5.2 1 1.3-1 1-3.3-.7-5.2-1.6-1.6-3.9-2.3-5.2-1m-10.8-8.1c-.7 1.3.3 2.9 2.3 3.9 1.6 1 3.6.7 4.3-.7.7-1.3-.3-2.9-2.3-3.9-2-.6-3.6-.3-4.3.7m32.4 35.6c-1.6 1.3-1 4.3 1.3 6.2 2.3 2.3 5.2 2.6 6.5 1 1.3-1.3.7-4.3-1.3-6.2-2.2-2.3-5.2-2.6-6.5-1m-11.4-14.7c-1.6 1-1.6 3.6 0 5.9s4.3 3.3 5.6 2.3c1.6-1.3 1.6-3.9 0-6.2-1.4-2.3-4-3.3-5.6-2" / > < / svg >
2024-10-24 15:00:38 +00:00
< / div >
< div class = "md-source__repository" >
PaddlePaddle/PaddleOCR
< / div >
< / a >
< / div >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "index.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Home
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "quick_start.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
快速开始
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "update.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
近期更新
< / span >
< / a >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_4" >
< label class = "md-nav__link" for = "__nav_4" id = "__nav_4_label" tabindex = "0" >
< span class = "md-ellipsis" >
低代码全流程开发
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_4_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_4" >
< span class = "md-nav__icon md-icon" > < / span >
低代码全流程开发
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "paddlex/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "paddlex/quick_start.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
快速开始
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_5" >
< div class = "md-nav__link md-nav__container" >
< a href = "model/index.html" class = "md-nav__link " >
< span class = "md-ellipsis" >
模型
< / span >
< / a >
< label class = "md-nav__link " for = "__nav_5" id = "__nav_5_label" tabindex = "0" >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< / div >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_5_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_5" >
< span class = "md-nav__icon md-icon" > < / span >
模型
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_5_2" >
< label class = "md-nav__link" for = "__nav_5_2" id = "__nav_5_2_label" tabindex = "0" >
< span class = "md-ellipsis" >
多硬件安装飞桨
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_5_2_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_5_2" >
< span class = "md-nav__icon md-icon" > < / span >
多硬件安装飞桨
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "model/hardware/install_other_devices.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
多硬件安装飞桨
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "model/hardware/supported_models.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
支持硬件列表
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_6" >
< label class = "md-nav__link" for = "__nav_6" id = "__nav_6_label" tabindex = "0" >
< span class = "md-ellipsis" >
PP-OCR 文本检测识别
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_6_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_6" >
< span class = "md-nav__icon md-icon" > < / span >
PP-OCR 文本检测识别
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppocr/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/quick_start.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
快速开始
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/installation.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
快速安装
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/visualization.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
效果展示
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/environment.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
运行环境
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_list.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
模型库
< / span >
< / a >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_6_7" >
< label class = "md-nav__link" for = "__nav_6_7" id = "__nav_6_7_label" tabindex = "0" >
< span class = "md-ellipsis" >
模型训练
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_6_7_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_6_7" >
< span class = "md-nav__icon md-icon" > < / span >
模型训练
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/training.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基本概念
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/detection.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
文本检测
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/recognition.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
文本识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/angle_class.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
文本方向分类器
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/kie.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
关键信息提取
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_train/finetune.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
模型微调
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_6_8" >
< label class = "md-nav__link" for = "__nav_6_8" id = "__nav_6_8_label" tabindex = "0" >
< span class = "md-ellipsis" >
模型压缩
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_6_8_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_6_8" >
< span class = "md-nav__icon md-icon" > < / span >
模型压缩
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppocr/model_compress/quantization.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
模型量化
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_compress/prune.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
模型裁剪
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/model_compress/knowledge_distillation.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
知识蒸馏
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_6_9" >
< div class = "md-nav__link md-nav__container" >
< a href = "ppocr/infer_deploy/index.html" class = "md-nav__link " >
< span class = "md-ellipsis" >
推理部署
< / span >
< / a >
< label class = "md-nav__link " for = "__nav_6_9" id = "__nav_6_9_label" tabindex = "0" >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< / div >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_6_9_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_6_9" >
< span class = "md-nav__icon md-icon" > < / span >
推理部署
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/python_infer.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基于Python预测引擎推理
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/cpp_infer.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基于C++预测引擎推理
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/windows_vs2019_build.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Visual Studio 2019 Community CMake 编译指南
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/paddle_server.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
服务化部署
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/android_demo.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Android部署
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/Jetson_infer.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Jetson部署
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/lite.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
端侧部署
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/paddle_js.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
网页前端部署
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/paddle2onnx.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Paddle2ONNX模型转化与预测
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/paddle_cloud.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
云上飞桨部署工具
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/infer_deploy/benchmark.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Benchmark
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_6_10" >
< label class = "md-nav__link" for = "__nav_6_10" id = "__nav_6_10_label" tabindex = "0" >
< span class = "md-ellipsis" >
博客
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_6_10_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_6_10" >
< span class = "md-nav__icon md-icon" > < / span >
博客
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppocr/blog/PP-OCRv3_introduction.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PP-OCRv3技术报告
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/PP-OCRv4_introduction.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PP-OCRv4技术报告
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/whl.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
paddleocr package使用说明
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/multi_languages.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
多语言模型
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/ocr_book.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
动手学OCR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/enhanced_ctc_loss.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Enhanced CTC Loss
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/slice.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
切片操作
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/inference_args.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PaddleOCR模型推理参数解释
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/distributed_training.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
分布式训练
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/clone.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
项目克隆
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/config.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
配置文件内容与生成
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppocr/blog/customize.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
如何生产自定义超轻量模型？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_7" >
< label class = "md-nav__link" for = "__nav_7" id = "__nav_7_label" tabindex = "0" >
< span class = "md-ellipsis" >
PP-Structure文档分析
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_7_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_7" >
< span class = "md-nav__icon md-icon" > < / span >
PP-Structure文档分析
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppstructure/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/quick_start.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
快速开始
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/models_list.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
模型库
< / span >
< / a >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_7_4" >
< label class = "md-nav__link" for = "__nav_7_4" id = "__nav_7_4_label" tabindex = "0" >
< span class = "md-ellipsis" >
模型训练
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_7_4_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_7_4" >
< span class = "md-nav__icon md-icon" > < / span >
模型训练
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppstructure/model_train/training.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基本概念
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/model_train/train_layout.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
版面分析
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/model_train/train_table.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
表格识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/model_train/recovery_to_doc.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
版面恢复
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/model_train/train_kie.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
关键信息提取
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_7_5" >
< div class = "md-nav__link md-nav__container" >
< a href = "ppstructure/infer_deploy/index.html" class = "md-nav__link " >
< span class = "md-ellipsis" >
推理部署
< / span >
< / a >
< label class = "md-nav__link " for = "__nav_7_5" id = "__nav_7_5_label" tabindex = "0" >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< / div >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_7_5_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_7_5" >
< span class = "md-nav__icon md-icon" > < / span >
推理部署
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppstructure/infer_deploy/python_infer.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基于Python预测引擎推理
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/infer_deploy/cpp_infer.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
基于C++预测引擎推理
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/infer_deploy/paddle_server.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
服务化部署
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_7_6" >
< label class = "md-nav__link" for = "__nav_7_6" id = "__nav_7_6_label" tabindex = "0" >
< span class = "md-ellipsis" >
博客
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_7_6_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_7_6" >
< span class = "md-nav__icon md-icon" > < / span >
博客
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "ppstructure/blog/return_word_pos.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
返回识别位置
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "ppstructure/blog/how_to_do_kie.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
怎样完成基于图像数据的信息抽取任务
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8" >
< label class = "md-nav__link" for = "__nav_8" id = "__nav_8_label" tabindex = "0" >
< span class = "md-ellipsis" >
前沿算法与模型
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_8_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8" >
< span class = "md-nav__icon md-icon" > < / span >
前沿算法与模型
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_2" >
< label class = "md-nav__link" for = "__nav_8_2" id = "__nav_8_2_label" tabindex = "0" >
< span class = "md-ellipsis" >
文本检测算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_2_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_2" >
< span class = "md-nav__icon md-icon" > < / span >
文本检测算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_db.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
DB与DB++
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_east.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
EAST
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_sast.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SAST
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_psenet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PSENet
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_fcenet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
FCENet
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_drrg.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
DRRG
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_detection/algorithm_det_ct.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
CT
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_3" >
< label class = "md-nav__link" for = "__nav_8_3" id = "__nav_8_3_label" tabindex = "0" >
< span class = "md-ellipsis" >
文本识别算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_3_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_3" >
< span class = "md-nav__icon md-icon" > < / span >
文本识别算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_crnn.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
CRNN
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_rosetta.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Rosetta
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_starnet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
STAR-Net
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_rare.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
RARE
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_srn.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SRN
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_nrtr.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
NRTR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_sar.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SAR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_seed.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SEED
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_svtr.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SVTR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_svtrv2.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SVTRv2
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_vitstr.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
ViTSTR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_abinet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
ABINet
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_visionlan.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
VisionLAN
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_spin.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SPIN
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_robustscanner.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
RobustScanner
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_rfl.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
RFL
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_parseq.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
ParseQ
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_cppd.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
CPPD
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/text_recognition/algorithm_rec_satrn.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SATRN
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_4" >
< label class = "md-nav__link" for = "__nav_8_4" id = "__nav_8_4_label" tabindex = "0" >
< span class = "md-ellipsis" >
文本超分辨率算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_4_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_4" >
< span class = "md-nav__icon md-icon" > < / span >
文本超分辨率算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/super_resolution/algorithm_sr_gestalt.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Text Gestalt
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/super_resolution/algorithm_sr_telescope.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
Text Telescope
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_5" >
< label class = "md-nav__link" for = "__nav_8_5" id = "__nav_8_5_label" tabindex = "0" >
< span class = "md-ellipsis" >
公式识别算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_5_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_5" >
< span class = "md-nav__icon md-icon" > < / span >
公式识别算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/formula_recognition/algorithm_rec_can.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
CAN
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/formula_recognition/algorithm_rec_latex_ocr.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
LaTeX-OCR
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_6" >
< label class = "md-nav__link" for = "__nav_8_6" id = "__nav_8_6_label" tabindex = "0" >
< span class = "md-ellipsis" >
端到端OCR算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_6_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_6" >
< span class = "md-nav__icon md-icon" > < / span >
端到端OCR算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/end_to_end/algorithm_e2e_pgnet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PGNet
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_7" >
< label class = "md-nav__link" for = "__nav_8_7" id = "__nav_8_7_label" tabindex = "0" >
< span class = "md-ellipsis" >
表格识别算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_7_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_7" >
< span class = "md-nav__icon md-icon" > < / span >
表格识别算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/table_recognition/algorithm_table_master.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
TableMaster
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/table_recognition/algorithm_table_slanet.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
TableSLANet
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_8_8" >
< label class = "md-nav__link" for = "__nav_8_8" id = "__nav_8_8_label" tabindex = "0" >
< span class = "md-ellipsis" >
关键信息抽取算法
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_8_8_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_8_8" >
< span class = "md-nav__icon md-icon" > < / span >
关键信息抽取算法
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "algorithm/kie/algorithm_kie_vi_layoutxlm.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
VI-LayoutXLM
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/kie/algorithm_kie_layoutxlm.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
LayoutLM
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/kie/algorithm_kie_sdmgr.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
SDMGR
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "algorithm/add_new_algorithm.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
使用PaddleOCR架构添加新算法
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_9" >
< label class = "md-nav__link" for = "__nav_9" id = "__nav_9_label" tabindex = "0" >
< span class = "md-ellipsis" >
场景应用
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_9_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_9" >
< span class = "md-nav__icon md-icon" > < / span >
场景应用
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "applications/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_9_2" >
< label class = "md-nav__link" for = "__nav_9_2" id = "__nav_9_2_label" tabindex = "0" >
< span class = "md-ellipsis" >
通用
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_9_2_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_9_2" >
< span class = "md-nav__icon md-icon" > < / span >
通用
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "applications/%E9%AB%98%E7%B2%BE%E5%BA%A6%E4%B8%AD%E6%96%87%E8%AF%86%E5%88%AB%E6%A8%A1%E5%9E%8B.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
高精度中文场景文本识别模型SVTR
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E6%89%8B%E5%86%99%E6%96%87%E5%AD%97%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
手写体识别
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_9_3" >
< label class = "md-nav__link" for = "__nav_9_3" id = "__nav_9_3_label" tabindex = "0" >
< span class = "md-ellipsis" >
制造
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_9_3_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_9_3" >
< span class = "md-nav__icon md-icon" > < / span >
制造
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "applications/%E5%85%89%E5%8A%9F%E7%8E%87%E8%AE%A1%E6%95%B0%E7%A0%81%E7%AE%A1%E5%AD%97%E7%AC%A6%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
数码管识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E6%B6%B2%E6%99%B6%E5%B1%8F%E8%AF%BB%E6%95%B0%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
液晶屏读数识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E5%8C%85%E8%A3%85%E7%94%9F%E4%BA%A7%E6%97%A5%E6%9C%9F%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
包装生产日期
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/PCB%E5%AD%97%E7%AC%A6%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
PCB文字识别
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_9_4" >
< label class = "md-nav__link" for = "__nav_9_4" id = "__nav_9_4_label" tabindex = "0" >
< span class = "md-ellipsis" >
金融
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_9_4_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_9_4" >
< span class = "md-nav__icon md-icon" > < / span >
金融
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "applications/%E5%A4%9A%E6%A8%A1%E6%80%81%E8%A1%A8%E5%8D%95%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
表单VQA
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E5%8F%91%E7%A5%A8%E5%85%B3%E9%94%AE%E4%BF%A1%E6%81%AF%E6%8A%BD%E5%8F%96.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
增值税发票
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E5%8D%B0%E7%AB%A0%E5%BC%AF%E6%9B%B2%E6%96%87%E5%AD%97%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
印章检测与识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E5%BF%AB%E9%80%9F%E6%9E%84%E5%BB%BA%E5%8D%A1%E8%AF%81%E7%B1%BBOCR.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
通用卡证识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "applications/%E6%89%AB%E6%8F%8F%E5%90%88%E5%90%8C%E5%85%B3%E9%94%AE%E4%BF%A1%E6%81%AF%E6%8F%90%E5%8F%96.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
合同比对
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_9_5" >
< label class = "md-nav__link" for = "__nav_9_5" id = "__nav_9_5_label" tabindex = "0" >
< span class = "md-ellipsis" >
交通
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "2" aria-labelledby = "__nav_9_5_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_9_5" >
< span class = "md-nav__icon md-icon" > < / span >
交通
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "applications/%E8%BD%BB%E9%87%8F%E7%BA%A7%E8%BD%A6%E7%89%8C%E8%AF%86%E5%88%AB.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
车牌识别
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_10" >
< label class = "md-nav__link" for = "__nav_10" id = "__nav_10_label" tabindex = "0" >
< span class = "md-ellipsis" >
数据标注与合成
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_10_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_10" >
< span class = "md-nav__icon md-icon" > < / span >
数据标注与合成
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "data_anno_synth/overview.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
概述
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "data_anno_synth/data_annotation.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
其它数据标注工具
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "data_anno_synth/data_synthesis.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
其它数据合成工具
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_11" >
< label class = "md-nav__link" for = "__nav_11" id = "__nav_11_label" tabindex = "0" >
< span class = "md-ellipsis" >
数据集
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_11_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_11" >
< span class = "md-nav__icon md-icon" > < / span >
数据集
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "datasets/datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
通用中英文OCR数据集
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "datasets/handwritten_datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
手写中文OCR数据集
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "datasets/vertical_and_multilingual_datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
垂类多语言OCR数据集
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "datasets/layout_datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
版面分析数据集
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "datasets/table_datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
表格识别数据集
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "datasets/kie_datasets.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
关键信息提取数据集
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--active" >
< input class = "md-nav__toggle md-toggle" type = "checkbox" id = "__toc" >
< label class = "md-nav__link md-nav__link--active" for = "__toc" >
< span class = "md-ellipsis" >
FAQ
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< a href = "FAQ.html" class = "md-nav__link md-nav__link--active" >
< span class = "md-ellipsis" >
FAQ
< / span >
< / a >
< nav class = "md-nav md-nav--secondary" aria-label = "目录" >
< label class = "md-nav__title" for = "__toc" >
< span class = "md-nav__icon md-icon" > < / span >
目录
< / label >
< ul class = "md-nav__list" data-md-component = "toc" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "#1" class = "md-nav__link" >
< span class = "md-ellipsis" >
1. 通用问题
< / span >
< / a >
< nav class = "md-nav" aria-label = "1. 通用问题" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#11" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.1 检测
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.1 检测" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 基于深度学习的文字检测方法有哪几种？各有什么优缺点？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#12" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.2 识别
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.2 识别" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR提供的文本识别算法包括哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-crnn" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 文本识别方法CRNN关键技术有哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ctcattention" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 对于中文行文本识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-tps" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 弯曲形变的文字识别需要怎么处理？ ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#13" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.3 端到端
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.3 端到端" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pgnetdbcrnnpgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 请问端到端的pgnet相比于DB+CRNN在准确率上有优势吗？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前OCR普遍是二阶段， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 二阶段的端到端的场景文本识别方法的不足有哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-aaai-2021pgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: AAAI 2021最新的端到端场景文本识别PGNet算法有什么特点？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#14" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.4 评估方法
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.4 评估方法" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-ocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: OCR领域常用的评估指标是什么？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#15" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.5 垂类场景实现思路
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.5 垂类场景实现思路" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ， ） ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_5" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_6" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_7" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qdb" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_8" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 弯曲文本（如略微形变的文档图像）漏检问题
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_9" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_10" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qopencvtps" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_11" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何识别招牌或者广告图中的艺术字？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_12" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 印章如何识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_13" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用预训练模型进行预测，对于特定字符识别识别效果较差，怎么解决？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_14" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在使用训练好的识别模型进行预测的时候，发现有很多重复的字，这个怎么解决呢？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ok90" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 图像正常识别出来的文字是OK的， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_15" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何识别竹简上的古文？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_16" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 只想要识别票据中的部分片段，重新训练它的话，只需要训练文本检测模型就可以了吗？问文本识别，方向分类还是用原来的模型这样可以吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何用PaddleOCR识别视频中的文字？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_17" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 相机采集的图像为四通道，应该如何处理？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_18" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 遇到中英文识别模型不支持的字符，该如何对模型做微调？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_19" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ） ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_20" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ） ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_21" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 多语言的字典里是混合了不同的语种，这个是有什么讲究吗？统一到一个字典里会对精度造成多大的损失？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_22" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ppocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 想把简历上的文字识别出来后， ， ， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#16" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.6 训练过程与模型调优
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.6 训练过程与模型调优" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-batch_size" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 增大batch_size模型训练速度没有明显提升
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_23" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预测时提示图像过大，显存、内存溢出了，应该如何处理？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-9070" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 识别训练时， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#17" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.7 补充资料
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.7 补充资料" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-ocr_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 对于小白如何快速入门中文OCR项目实践？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#2-paddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
2. PaddleOCR实战问题
< / span >
< / a >
< nav class = "md-nav" aria-label = "2. PaddleOCR实战问题" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#21-paddleocr-repo" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.1 PaddleOCR repo
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.1 PaddleOCR repo" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocr-developdygraph" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR develop分支和dygraph分支的区别？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#22" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.2 安装环境
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.2 安装环境" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qoserror-winerror-126-mac-pro-python-34-shapely-import" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddlepaddlegpu-osenvironcuda_visible_devices" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrwindowsmac" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#23" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.3 数据量说明
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.3 数据量说明" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrgpuepoch" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？ ， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q30w500w" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_24" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 当训练数据量少时，如何获取更多的数据？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#24" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.4 数据标注与生成
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.4 数据标注与生成" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-style-text" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: Style-Text 如何不文字风格迁移，就像普通文本生成程序一样默认字体直接输出到分割的背景图？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 能否修改StyleText配置文件中的分辨率？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText是否可以更换字体文件？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText批量生成图片为什么没有输出？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qstyletexttextinputstyleinput" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText 合成数据效果不好？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#25" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.5 预训练模型与微调
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.5 预训练模型与微调" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qbackbone" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_25" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 参照文档做实际项目时，是重新训练还是在官方训练的基础上进行训练？具体如何操作？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-inferencepdiparams-inferencepdmodel" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 下载的识别模型解压后缺失文件，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-checkpointsload" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么在checkpoints中load下载的预训练模型会报错？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-finetune" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何对检测模型finetune， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#26" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.6 模型超参调整
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.6 模型超参调整" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-db640" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB检测训练输入尺寸640， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-32" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预处理部分， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-stride2-1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在识别模型中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qshape" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_26" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ，
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#27" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.7 模型结构
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.7 模型结构" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qlstm" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qlstmgru" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qcrnnbackbonedensenetresnet_vd" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-backbone" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何根据不同的硬件平台选用不同的backbone？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#28-pp-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.8 PP-OCR系统
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.8 PP-OCR系统" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pp-ocrse" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在PP-OCR系统中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pp-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PP-OCR系统中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-db" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB文本检测， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_27" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前知识蒸馏有哪些主要的实践思路？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-preds_idx-predsargmaxaxis2beam-search" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 文字识别模型模型的输出矩阵需要进行解码才能得到识别的文本。代码中实现为preds_idx = preds.argmax(axis=2)， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#29" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.9 端到端
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.9 端到端" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 端到端算法PGNet是否支持中文识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 端到端算法PGNet提供了两种后处理方式， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgneteval" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用PGNet进行eval报错？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PGNet有中文预训练模型吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 用于PGNet的训练集， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 用PGNet做进行端到端训练时，
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#210" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.10 模型效果与效果不一致
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.10 模型效果与效果不一致" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pp-ocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PP-OCR检测效果不好， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q2126" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-db_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB有些框太贴文本了反而去掉了一些文本的边角影响识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-infer" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用合成数据精调小模型后， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_28" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 表格识别中，如何提高单字的识别结果？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-dygraphrelease20" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 动态图分支(dygraph,release/2.0)，训练模型和推理模型效果不一致
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-detinferenceeval" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 自己训练的det模型， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_29" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练模型和测试模型的检测结果差距较大
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrpythonc" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR模型Python端预测和C++预测结果不一致？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#211" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.11 训练调试与配置文件
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.11 训练调试与配置文件" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-epoch" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 某个类别的样本比较少， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_30" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_31" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_32" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-config-ymlratio_list" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: config yml文件中的ratio_list参数的作用是什么？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-iaa" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: iaa里面添加的数据增强方式， ？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_33" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 怎么加速训练过程呢？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-finetune_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 一些特殊场景的数据识别效果差， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR可以识别灰度图吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_34" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何合成手写中文数据集？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocr200step" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrcosine_decay" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR在训练的时候一直使用cosine_decay的学习率下降策略， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-cosine" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: Cosine学习率的更新策略是怎样的？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-cosinewarmup" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 之前的CosineWarmup方法为什么不见了？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-warmup" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练识别和检测时学习率要加上warmup， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-dygraph" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 关于dygraph分支中， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_35" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练过程中，训练程序意外退出/挂起，应该如何解决？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-log" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练程序启动后直到结束， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-num-workers" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 配置文件中的参数num workers是什么意思， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#212" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.12 预测
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.12 预测" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocrtest_batch_size_per_card1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么PaddleOCR检测预测是只支持一张图片测试？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrtensorrt" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR支持tensorrt推理吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-tensorrtpaddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何使用TensorRT加速PaddleOCR预测？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_36" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么识别模型做预测的时候，预测图片的数量数量还会影响预测的精度
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#213" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.13 推理部署
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.13 推理部署" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qpaddleocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrcputenorrtgpu" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qhubservingpdserving" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddle-hub-serving-imgpathimgurl" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前paddle hub serving 只支持 imgpath，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-c-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: C++ 端侧部署可以只对OCR的检测部署吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_37" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-litenb" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: lite预测库和nb模型版本不匹配， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrsdk" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrtest_batch_size_per_card1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_38" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddle-lite" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 采用Paddle-Lite进行端侧部署， ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何多进程运行paddleocr？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_39" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何多进程预测？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrt4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 怎么解决paddleOCR在T4卡上有越预测越慢的情况？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-windowscpp-inferencepaddle_fluiddllopencv_world346dll" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在windows上进行cpp inference的部署时，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-winc" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: win下C++部署中文识别乱码的解决方法
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-windows-3060gpu" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: windows 3060显卡GPU模式启动 加载模型慢
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qmac" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_40" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-hub_serving" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用hub_serving部署， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddlelite" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在使用PaddleLite进行预测部署时，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_41" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预测时显存爆炸、内存泄漏问题？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item md-nav__item--nested" >
< input class = "md-nav__toggle md-toggle md-toggle--indeterminate" type = "checkbox" id = "__nav_13" >
< label class = "md-nav__link" for = "__nav_13" id = "__nav_13_label" tabindex = "0" >
< span class = "md-ellipsis" >
社区
< / span >
< span class = "md-nav__icon md-icon" > < / span >
< / label >
< nav class = "md-nav" data-md-level = "1" aria-labelledby = "__nav_13_label" aria-expanded = "false" >
< label class = "md-nav__title" for = "__nav_13" >
< span class = "md-nav__icon md-icon" > < / span >
社区
< / label >
< ul class = "md-nav__list" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "community/community_contribution.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
社区贡献
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "community/code_and_doc.html" class = "md-nav__link" >
< span class = "md-ellipsis" >
附录
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / div >
< / div >
< / div >
< div class = "md-sidebar md-sidebar--secondary" data-md-component = "sidebar" data-md-type = "toc" hidden >
< div class = "md-sidebar__scrollwrap" >
< div class = "md-sidebar__inner" >
< nav class = "md-nav md-nav--secondary" aria-label = "目录" >
< label class = "md-nav__title" for = "__toc" >
< span class = "md-nav__icon md-icon" > < / span >
目录
< / label >
< ul class = "md-nav__list" data-md-component = "toc" data-md-scrollfix >
< li class = "md-nav__item" >
< a href = "#1" class = "md-nav__link" >
< span class = "md-ellipsis" >
1. 通用问题
< / span >
< / a >
< nav class = "md-nav" aria-label = "1. 通用问题" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#11" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.1 检测
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.1 检测" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 基于深度学习的文字检测方法有哪几种？各有什么优缺点？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#12" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.2 识别
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.2 识别" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR提供的文本识别算法包括哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-crnn" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 文本识别方法CRNN关键技术有哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ctcattention" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 对于中文行文本识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-tps" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 弯曲形变的文字识别需要怎么处理？ ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#13" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.3 端到端
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.3 端到端" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pgnetdbcrnnpgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 请问端到端的pgnet相比于DB+CRNN在准确率上有优势吗？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前OCR普遍是二阶段， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 二阶段的端到端的场景文本识别方法的不足有哪些？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-aaai-2021pgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: AAAI 2021最新的端到端场景文本识别PGNet算法有什么特点？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#14" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.4 评估方法
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.4 评估方法" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-ocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: OCR领域常用的评估指标是什么？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#15" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.5 垂类场景实现思路
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.5 垂类场景实现思路" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ， ） ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_5" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_6" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_7" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qdb" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_8" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 弯曲文本（如略微形变的文档图像）漏检问题
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_9" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_10" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qopencvtps" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_11" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何识别招牌或者广告图中的艺术字？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_12" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 印章如何识别
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_13" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用预训练模型进行预测，对于特定字符识别识别效果较差，怎么解决？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_14" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在使用训练好的识别模型进行预测的时候，发现有很多重复的字，这个怎么解决呢？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ok90" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 图像正常识别出来的文字是OK的， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_15" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何识别竹简上的古文？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_16" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 只想要识别票据中的部分片段，重新训练它的话，只需要训练文本检测模型就可以了吗？问文本识别，方向分类还是用原来的模型这样可以吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何用PaddleOCR识别视频中的文字？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_17" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 相机采集的图像为四通道，应该如何处理？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_18" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 遇到中英文识别模型不支持的字符，该如何对模型做微调？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_19" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ） ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_20" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： （ ） ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_21" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 多语言的字典里是混合了不同的语种，这个是有什么讲究吗？统一到一个字典里会对精度造成多大的损失？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_22" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-ppocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 想把简历上的文字识别出来后， ， ， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#16" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.6 训练过程与模型调优
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.6 训练过程与模型调优" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-batch_size" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 增大batch_size模型训练速度没有明显提升
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_23" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预测时提示图像过大，显存、内存溢出了，应该如何处理？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-9070" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 识别训练时， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#17" class = "md-nav__link" >
< span class = "md-ellipsis" >
1.7 补充资料
< / span >
< / a >
< nav class = "md-nav" aria-label = "1.7 补充资料" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-ocr_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 对于小白如何快速入门中文OCR项目实践？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#2-paddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
2. PaddleOCR实战问题
< / span >
< / a >
< nav class = "md-nav" aria-label = "2. PaddleOCR实战问题" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#21-paddleocr-repo" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.1 PaddleOCR repo
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.1 PaddleOCR repo" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocr-developdygraph" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR develop分支和dygraph分支的区别？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#22" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.2 安装环境
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.2 安装环境" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qoserror-winerror-126-mac-pro-python-34-shapely-import" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddlepaddlegpu-osenvironcuda_visible_devices" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrwindowsmac" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#23" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.3 数据量说明
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.3 数据量说明" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrgpuepoch" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？ ， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q30w500w" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_24" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 当训练数据量少时，如何获取更多的数据？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#24" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.4 数据标注与生成
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.4 数据标注与生成" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-style-text" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: Style-Text 如何不文字风格迁移，就像普通文本生成程序一样默认字体直接输出到分割的背景图？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 能否修改StyleText配置文件中的分辨率？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText是否可以更换字体文件？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText批量生成图片为什么没有输出？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qstyletexttextinputstyleinput" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-styletext_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: StyleText 合成数据效果不好？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#25" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.5 预训练模型与微调
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.5 预训练模型与微调" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qbackbone" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_25" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 参照文档做实际项目时，是重新训练还是在官方训练的基础上进行训练？具体如何操作？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-inferencepdiparams-inferencepdmodel" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 下载的识别模型解压后缺失文件，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-checkpointsload" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么在checkpoints中load下载的预训练模型会报错？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-finetune" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何对检测模型finetune， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#26" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.6 模型超参调整
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.6 模型超参调整" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-db640" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB检测训练输入尺寸640， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-32" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预处理部分， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-stride2-1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在识别模型中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qshape" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_26" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ，
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#27" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.7 模型结构
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.7 模型结构" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qlstm" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qlstmgru" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qcrnnbackbonedensenetresnet_vd" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-backbone" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何根据不同的硬件平台选用不同的backbone？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#28-pp-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.8 PP-OCR系统
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.8 PP-OCR系统" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pp-ocrse" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在PP-OCR系统中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pp-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PP-OCR系统中， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-db" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB文本检测， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_27" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前知识蒸馏有哪些主要的实践思路？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-preds_idx-predsargmaxaxis2beam-search" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 文字识别模型模型的输出矩阵需要进行解码才能得到识别的文本。代码中实现为preds_idx = preds.argmax(axis=2)， ， ， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#29" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.9 端到端
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.9 端到端" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pgnet" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 端到端算法PGNet是否支持中文识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 端到端算法PGNet提供了两种后处理方式， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgneteval" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用PGNet进行eval报错？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PGNet有中文预训练模型吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 用于PGNet的训练集， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-pgnet_4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 用PGNet做进行端到端训练时，
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#210" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.10 模型效果与效果不一致
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.10 模型效果与效果不一致" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-pp-ocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PP-OCR检测效果不好， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q2126" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-db_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: DB有些框太贴文本了反而去掉了一些文本的边角影响识别， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-infer" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用合成数据精调小模型后， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_28" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 表格识别中，如何提高单字的识别结果？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-dygraphrelease20" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 动态图分支(dygraph,release/2.0)，训练模型和推理模型效果不一致
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-detinferenceeval" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 自己训练的det模型， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_29" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练模型和测试模型的检测结果差距较大
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrpythonc" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR模型Python端预测和C++预测结果不一致？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#211" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.11 训练调试与配置文件
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.11 训练调试与配置文件" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-epoch" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 某个类别的样本比较少， ， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_30" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_31" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_32" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-config-ymlratio_list" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: config yml文件中的ratio_list参数的作用是什么？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-iaa" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: iaa里面添加的数据增强方式， ？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_33" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 怎么加速训练过程呢？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-finetune_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 一些特殊场景的数据识别效果差， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_2" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR可以识别灰度图吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_34" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何合成手写中文数据集？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocr200step" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrcosine_decay" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR在训练的时候一直使用cosine_decay的学习率下降策略， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-cosine" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: Cosine学习率的更新策略是怎样的？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-cosinewarmup" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 之前的CosineWarmup方法为什么不见了？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-warmup" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练识别和检测时学习率要加上warmup， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-dygraph" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 关于dygraph分支中， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_35" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练过程中，训练程序意外退出/挂起，应该如何解决？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-log" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 训练程序启动后直到结束， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-num-workers" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 配置文件中的参数num workers是什么意思， ？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#212" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.12 预测
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.12 预测" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#q-paddleocrtest_batch_size_per_card1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么PaddleOCR检测预测是只支持一张图片测试？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrtensorrt" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: PaddleOCR支持tensorrt推理吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-tensorrtpaddleocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何使用TensorRT加速PaddleOCR预测？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_36" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 为什么识别模型做预测的时候，预测图片的数量数量还会影响预测的精度
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< li class = "md-nav__item" >
< a href = "#213" class = "md-nav__link" >
< span class = "md-ellipsis" >
2.13 推理部署
< / span >
< / a >
< nav class = "md-nav" aria-label = "2.13 推理部署" >
< ul class = "md-nav__list" >
< li class = "md-nav__item" >
< a href = "#qpaddleocr_1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrcputenorrtgpu" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ， ？ ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qhubservingpdserving" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddle-hub-serving-imgpathimgurl" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 目前paddle hub serving 只支持 imgpath，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-c-ocr" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: C++ 端侧部署可以只对OCR的检测部署吗？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_37" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-litenb" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: lite预测库和nb模型版本不匹配， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrsdk" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q：
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qpaddleocrtest_batch_size_per_card1" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_38" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddle-lite" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 采用Paddle-Lite进行端侧部署， ，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocr_3" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何多进程运行paddleocr？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_39" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 如何多进程预测？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddleocrt4" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 怎么解决paddleOCR在T4卡上有越预测越慢的情况？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-windowscpp-inferencepaddle_fluiddllopencv_world346dll" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在windows上进行cpp inference的部署时，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-winc" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: win下C++部署中文识别乱码的解决方法
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-windows-3060gpu" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: windows 3060显卡GPU模式启动 加载模型慢
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#qmac" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_40" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q： ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-hub_serving" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 使用hub_serving部署， ， ？
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q-paddlelite" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 在使用PaddleLite进行预测部署时，
< / span >
< / a >
< / li >
< li class = "md-nav__item" >
< a href = "#q_41" class = "md-nav__link" >
< span class = "md-ellipsis" >
Q: 预测时显存爆炸、内存泄漏问题？
< / span >
< / a >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / li >
< / ul >
< / nav >
< / div >
< / div >
< / div >
< div class = "md-content" data-md-component = "content" >
< article class = "md-content__inner md-typeset" >
< a href = "https://github.com/PaddlePaddle/PaddleOCR/edit/main/docs/FAQ.md" title = "编辑此页" class = "md-content__button md-icon" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M20.71 7.04c.39-.39.39-1.04 0-1.41l-2.34-2.34c-.37-.39-1.02-.39-1.41 0l-1.84 1.83 3.75 3.75M3 17.25V21h3.75L17.81 9.93l-3.75-3.75z" / > < / svg >
< / a >
< a href = "https://github.com/PaddlePaddle/PaddleOCR/raw/main/docs/FAQ.md" title = "查看本页的源代码" class = "md-content__button md-icon" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M12 9a3 3 0 0 0-3 3 3 3 0 0 0 3 3 3 3 0 0 0 3-3 3 3 0 0 0-3-3m0 8a5 5 0 0 1-5-5 5 5 0 0 1 5-5 5 5 0 0 1 5 5 5 5 0 0 1-5 5m0-12.5C7 4.5 2.73 7.61 1 12c1.73 4.39 6 7.5 11 7.5s9.27-3.11 11-7.5c-1.73-4.39-6-7.5-11-7.5" / > < / svg >
< / a >
< h1 > FAQ< / h1 >
< blockquote >
< p > 恭喜你发现宝藏！< / p >
< / blockquote >
< p > PaddleOCR收集整理了自从开源以来在issues和用户群中的常见问题并且给出了简要解答， ， < / p >
< p > 其中< a href = "#1" > 通用问题< / a > 一般是初次接触OCR相关算法时用户会提出的问题， < a href = "#15" > 1.5 垂类场景实现思路< / a > 中总结了如何在一些具体的场景中确定技术路线进行优化。< a href = "#2-paddleocr" > PaddleOCR常见问题< / a > 是开发者在使用PaddleOCR之后可能会遇到的问题也是PaddleOCR实践过程中的避坑指南。< / p >
< p > 同时PaddleOCR也会在review issue的过程中添加 < code > good issue< / code > 、 < code > good first issue< / code > 标签， ， < / p >
< p > OCR领域大佬众多， ， ， ， < strong > 希望有识之士帮忙补充和修正< / strong > ，万分感谢。< / p >
< h2 id = "1" > 1. 通用问题< a class = "headerlink" href = "#1" title = "Permanent link" > ¶ < / a > < / h2 >
< h3 id = "11" > 1.1 检测< a class = "headerlink" href = "#11" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q" > Q: 基于深度学习的文字检测方法有哪几种？各有什么优缺点？< a class = "headerlink" href = "#q" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：常用的基于深度学习的文字检测方法一般可以分为基于回归的、基于分割的两大类，当然还有一些将两者进行结合的方法。< / p >
< p > （ ） ， ， ， < / p >
< p > （ ） ， ， ， ， ， ， ， ， ， ， ， < / p >
< h3 id = "12" > 1.2 识别< a class = "headerlink" href = "#12" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-paddleocr" > Q: PaddleOCR提供的文本识别算法包括哪些？ < a class = "headerlink" href = "#q-paddleocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ； ， ， < / p >
< h4 id = "q-crnn" > Q: 文本识别方法CRNN关键技术有哪些？ < a class = "headerlink" href = "#q-crnn" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） （ ） ， （ ） < / p >
< h4 id = "q-ctcattention" > Q: 对于中文行文本识别， ？ < a class = "headerlink" href = "#q-ctcattention" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ， ， ， ， ， ， < / p >
< p > （ ） ， ， ， < / p >
< h4 id = "q-tps" > Q: 弯曲形变的文字识别需要怎么处理？ ， ？ < a class = "headerlink" href = "#q-tps" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ， ， ， < / p >
< p > （ ） ， （ ） ， ， ， ， （ ， ） ， ， ， < / p >
< blockquote >
< p > < strong > Warning< / strong > ： ， ， ， < / p >
< / blockquote >
< h3 id = "13" > 1.3 端到端< a class = "headerlink" href = "#13" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-pgnetdbcrnnpgnet" > Q: 请问端到端的pgnet相比于DB+CRNN在准确率上有优势吗？ ？ < a class = "headerlink" href = "#q-pgnetdbcrnnpgnet" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ； ， ， < / p >
< h4 id = "q-ocr" > Q: 目前OCR普遍是二阶段， ？ < a class = "headerlink" href = "#q-ocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， < / p >
< h4 id = "q_1" > Q: 二阶段的端到端的场景文本识别方法的不足有哪些？< a class = "headerlink" href = "#q_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "q-aaai-2021pgnet" > Q: AAAI 2021最新的端到端场景文本识别PGNet算法有什么特点？ < a class = "headerlink" href = "#q-aaai-2021pgnet" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， < / p >
< h3 id = "14" > 1.4 评估方法< a class = "headerlink" href = "#14" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-ocr_1" > Q: OCR领域常用的评估指标是什么？ < a class = "headerlink" href = "#q-ocr_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：对于两阶段的可以分开来看，分别是检测和识别阶段< / p >
< p > （ ） ： ， ， ： ， ： ， < / p >
< p > （ ） ：
字符识别准确率，即正确识别的文本行占标注的文本行数量的比例，只有整行文本识别对才算正确识别。< / p >
< p > （ ） ：
端对端召回率：准确检测并正确识别文本行在全部标注文本行的占比；
端到端准确率：准确检测并正确识别文本行在 检测到的文本行数量 的占比；
准确检测的标准是检测框与标注框的IOU大于某个阈值， < / p >
< h3 id = "15" > 1.5 垂类场景实现思路< a class = "headerlink" href = "#15" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q_2" > Q： （ ， ） ， ？ < a class = "headerlink" href = "#q_2" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ， ， ， ， ； ， < / p >
< p > （ ） ， < / p >
< h4 id = "q_3" > Q： ， ？ < a class = "headerlink" href = "#q_3" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， ， < / p >
< h4 id = "q_4" > Q： < a class = "headerlink" href = "#q_4" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， < / p >
< h4 id = "q_5" > Q： ， ？ < a class = "headerlink" href = "#q_5" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， < / p >
< h4 id = "q_6" > Q： < a class = "headerlink" href = "#q_6" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ： ， ， < / p >
< h4 id = "q_7" > Q： ， ， ？ < a class = "headerlink" href = "#q_7" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ： < / p >
< h4 id = "qdb" > Q： ， ？ < a class = "headerlink" href = "#qdb" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：可以在预测时调小 det_db_box_thresh 阈值， < / p >
< h4 id = "q_8" > Q: 弯曲文本（如略微形变的文档图像）漏检问题< a class = "headerlink" href = "#q_8" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : db后处理中计算文本框平均得分时， ， ， ， ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/pull/2604" > 可视化对比效果< / a > 。该功能通过参数 < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/tools/infer/utility.py#L51" > det_db_score_mode< / a > 进行选择，参数值可选[< code > fast< / code > (默认)、< code > slow< / code > ]， < code > fast< / code > 对应原始的rectangle方式， < code > slow< / code > 对应polygon方式。感谢用户< a href = "https://github.com/buptlihang" > buptlihang< / a > 提< a href = "https://github.com/PaddlePaddle/PaddleOCR/pull/2574" > pr< / a > 帮助解决该问题🌹。< / p >
< h4 id = "q_9" > Q： ？ < a class = "headerlink" href = "#q_9" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：在中文识别模型训练时，并不是采用直接将训练样本缩放到[3,32,320]进行训练， ， ， ， ， ， ， ， ， ， < / p >
< h4 id = "q_10" > Q： ？ < a class = "headerlink" href = "#q_10" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：空格识别可以考虑以下两种方案：< / p >
< p > (1)优化文本检测算法。检测结果在空格处将文本断开。这种方案在检测数据标注时，需要将含有空格的文本行分成好多段。< / p >
< p > (2)优化文本识别算法。在识别字典里面引入空格字符，然后在识别的训练数据中，如果用空行，进行标注。此外，合成数据时，通过拼接训练数据，生成含有空格的文本。< / p >
< h4 id = "qopencvtps" > Q： ？ < a class = "headerlink" href = "#qopencvtps" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "q_11" > Q: 如何识别招牌或者广告图中的艺术字？< a class = "headerlink" href = "#q_11" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 招牌或者广告图中的艺术字是文本识别一个非常有挑战性的难题， ， ， ， < / p >
< h4 id = "q_12" > Q: 印章如何识别< a class = "headerlink" href = "#q_12" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< h4 id = "q_13" > Q: 使用预训练模型进行预测，对于特定字符识别识别效果较差，怎么解决？< a class = "headerlink" href = "#q_13" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 由于我们所提供的识别模型是基于通用大规模数据集进行训练的， ， ， ， ， ： < / p >
< h4 id = "q_14" > Q: 在使用训练好的识别模型进行预测的时候，发现有很多重复的字，这个怎么解决呢？< a class = "headerlink" href = "#q_14" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：可以看下训练的尺度和预测的尺度是否相同，如果训练的尺度为[3, 32, 320]，预测的尺度为[3, 64, 640]，则会有比较多的重复识别现象。< / p >
< h4 id = "q-ok90" > Q: 图像正常识别出来的文字是OK的， ， ？ < a class = "headerlink" href = "#q-ok90" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 整图旋转90之后效果变差是有可能的， ； < / p >
< h4 id = "q_15" > Q: 如何识别竹简上的古文？< a class = "headerlink" href = "#q_15" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < a href = "https://github.com/PFCCLab/StyleText" > StyleText< / a > 工具。
而如果使用的字符是特殊的古文字、甲骨文、象形文字等，那么首先需要构建一个古文字的字典，之后再进行训练。< / p >
< h4 id = "q_16" > Q: 只想要识别票据中的部分片段，重新训练它的话，只需要训练文本检测模型就可以了吗？问文本识别，方向分类还是用原来的模型这样可以吗？< a class = "headerlink" href = "#q_16" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "q-paddleocr_1" > Q: 如何用PaddleOCR识别视频中的文字？ < a class = "headerlink" href = "#q-paddleocr_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 目前PaddleOCR主要针对图像做处理， ， ， < / p >
< h4 id = "q_17" > Q: 相机采集的图像为四通道，应该如何处理？< a class = "headerlink" href = "#q_17" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 有两种方式处理：< / p >
< ul >
< li > 如果没有其他需要， ， ， < / li >
< li > 如果其他模块需要处理四通道的图像， ， & img,img3chan,CV_RGBA2RGB)。< / li >
< / ul >
< h4 id = "q_18" > Q: 遇到中英文识别模型不支持的字符，该如何对模型做微调？< a class = "headerlink" href = "#q_18" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：如果希望识别中英文识别模型中不支持的字符，需要更新识别的字典，并完成微调过程。比如说如果希望模型能够进一步识别罗马数字，可以按照以下步骤完成模型微调过程。< / p >
< ol >
< li > 准备中英文识别数据以及罗马数字的识别数据，用于训练，同时保证罗马数字和中英文识别数字的效果；< / li >
< li > 修改默认的字典文件，在后面添加罗马数字的字符；< / li >
< li > 下载PaddleOCR提供的预训练模型， ， < / li >
< / ol >
< h4 id = "q_19" > Q： （ ） ？ < a class = "headerlink" href = "#q_19" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：首先请您确认要识别的特殊字符是否在字典中。
如果字符在已经字典中但效果依然不好， ， < / p >
< hr / >
< h4 id = "q_20" > Q： （ ） ， ？ < a class = "headerlink" href = "#q_20" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ，
其中不同类型文本共用同一个检测模型， ， ， ， ， ， ， < / p >
< h4 id = "q_21" > Q: 多语言的字典里是混合了不同的语种，这个是有什么讲究吗？统一到一个字典里会对精度造成多大的损失？< a class = "headerlink" href = "#q_21" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， < / p >
< h4 id = "q_22" > Q： ， ， < a class = "headerlink" href = "#q_22" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：处理字符的时候，把多字符的当作一个字就行，字典中每行是一个字。< / p >
< hr / >
< h4 id = "q-ppocr" > Q: 想把简历上的文字识别出来后， ， ， ， ， ？ < a class = "headerlink" href = "#q-ppocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 这样的需求在企业应用中确实比较常见，但往往都是个性化的需求，没有非常规整统一的处理方式。常见的处理方式有如下两种：< / p >
< ol >
< li > 对于单一版式、或者版式差异不大的应用场景，可以基于识别场景的一些先验信息，将识别内容进行配对； 比如运用表单结构信息：常见表单"姓名"关键字的后面，往往紧跟的就是名字信息< / li >
< li > 对于版式多样，或者无固定版式的场景， 需要借助于NLP中的NER技术， ， < / li >
< / ol >
< p > 由于这部分需求和业务场景强相关， ， ， < a href = "https://github.com/PaddlePaddle/ERNIE" > PaddlePaddle/ERNIE< / a > ， < / p >
< h3 id = "16" > 1.6 训练过程与模型调优< a class = "headerlink" href = "#16" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-batch_size" > Q: 增大batch_size模型训练速度没有明显提升< a class = "headerlink" href = "#q-batch_size" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ：
export FLAGS_initial_cpu_memory_in_mb=2000 # 设置初始化内存约2G左右< / p >
< h4 id = "q_23" > Q: 预测时提示图像过大，显存、内存溢出了，应该如何处理？< a class = "headerlink" href = "#q_23" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< h4 id = "q-9070" > Q: 识别训练时， ， ， ？ < a class = "headerlink" href = "#q-9070" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ： （ ） ， （ ） < / p >
< h3 id = "17" > 1.7 补充资料< a class = "headerlink" href = "#17" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-ocr_2" > Q: 对于小白如何快速入门中文OCR项目实践？ < a class = "headerlink" href = "#q-ocr_2" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， ： < a href = "https://aistudio.baidu.com/aistudio/course/introduce/1519" > https://aistudio.baidu.com/aistudio/course/introduce/1519< / a > < / p >
< h2 id = "2-paddleocr" > 2. PaddleOCR实战问题< a class = "headerlink" href = "#2-paddleocr" title = "Permanent link" > ¶ < / a > < / h2 >
< h3 id = "21-paddleocr-repo" > 2.1 PaddleOCR repo< a class = "headerlink" href = "#21-paddleocr-repo" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-paddleocr-developdygraph" > Q: PaddleOCR develop分支和dygraph分支的区别？ < a class = "headerlink" href = "#q-paddleocr-developdygraph" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ： < / p >
< ul >
< li > develop： ， ， ， < / li >
< li > release/1.1： < / li >
< li > dygraph： ， ， ， < / li >
< li > release/2.0-rc1-0： ， < / li >
< / ul >
< p > 如果您已经上手过PaddleOCR， ， ， ， ， ， < / p >
< p > < strong > 注意< / strong > ： ， < / p >
< h4 id = "qpaddleocrocr" > Q： ？ < a class = "headerlink" href = "#qpaddleocrocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ，
如果缺少带标注的数据， ， ， < / p >
< h3 id = "22" > 2.2 安装环境< a class = "headerlink" href = "#22" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "qoserror-winerror-126-mac-pro-python-34-shapely-import" > Q： ： < a class = "headerlink" href = "#qoserror-winerror-126-mac-pro-python-34-shapely-import" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "qpaddlepaddlegpu-osenvironcuda_visible_devices" > Q： < a class = "headerlink" href = "#qpaddlepaddlegpu-osenvironcuda_visible_devices" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：通过设置 export CUDA_VISIBLE_DEVICES='0'环境变量< / p >
< h4 id = "qpaddleocrwindowsmac" > Q： ？ < a class = "headerlink" href = "#qpaddleocrwindowsmac" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ：
（ ） ， ， ，
（ ） ， ， ， < / p >
< h3 id = "23" > 2.3 数据量说明< a class = "headerlink" href = "#23" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "qocr" > Q： ， ？ < a class = "headerlink" href = "#qocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ， ， ， < / p >
< p > （ ） ， ， ， （ ， ， ， ） ， < / p >
< h4 id = "qpaddleocrgpuepoch" > Q： ？ ， ， ， ？ < a class = "headerlink" href = "#qpaddleocrgpuepoch" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：
（ ） ， ， ， ， ； ：
（ ） ， （ ） ， ： ， ： ， < / p >
< p > 超轻量模型训练分为2个阶段：
(1)全量数据训练50epoch，
(2)合成数据+真实数据按照1:1数据采样， ， < / p >
< p > 通用模型训练：
真实数据+合成数据，动态采样(1： ， ， < / p >
< h4 id = "q30w500w" > Q： ， ， ， ？ < a class = "headerlink" href = "#q30w500w" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ： ： ， ， ， ， < / p >
< h4 id = "q_24" > Q: 当训练数据量少时，如何获取更多的数据？< a class = "headerlink" href = "#q_24" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ： （ ） ， （ ） ， ， （ ） ， < / p >
< h3 id = "24" > 2.4 数据标注与生成< a class = "headerlink" href = "#24" title = "Permanent link" > ¶ < / a > < / h3 >
< blockquote >
< p > [!NOTE]
StyleText 已经移动到 < a href = "https://github.com/PFCCLab/StyleText" > PFCCLab/StyleText< / a > < / p >
< / blockquote >
< h4 id = "q-style-text" > Q: Style-Text 如何不文字风格迁移，就像普通文本生成程序一样默认字体直接输出到分割的背景图？< a class = "headerlink" href = "#q-style-text" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ：
< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/de3e2e7cd3b8b65ee02d7a41e570fa5b511a3c1d/StyleText/engine/synthesisers.py#L68" > https://github.com/PaddlePaddle/PaddleOCR/blob/de3e2e7cd3b8b65ee02d7a41e570fa5b511a3c1d/StyleText/engine/synthesisers.py#L68< / a > < / p >
< h4 id = "q-styletext" > Q: 能否修改StyleText配置文件中的分辨率？ < a class = "headerlink" href = "#q-styletext" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "q-styletext_1" > Q: StyleText是否可以更换字体文件？ < a class = "headerlink" href = "#q-styletext_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ，
StyleText的用途主要是： ， < / p >
< h4 id = "q-styletext_2" > Q: StyleText批量生成图片为什么没有输出？ < a class = "headerlink" href = "#q-styletext_2" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：需要检查以下您配置文件中的路径是否都存在。尤其要注意的是< a href = "https://github.com/PFCCLab/StyleText/blob/main/README_ch.md#%E4%B8%89%E5%BF%AB%E9%80%9F%E4%B8%8A%E6%89%8B" > label_file配置< / a > 。
如果您使用的style_image输入没有label信息， < / p >
< h4 id = "qstyletexttextinputstyleinput" > Q： ， ， ？ < a class = "headerlink" href = "#qstyletexttextinputstyleinput" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ： < / p >
< ol >
< li > 将StyleInput按列的方向进行复制与扩充， < / li >
< li > 将TextInput进行裁剪， ， < / li >
< / ol >
< p > 实际使用中发现， ， < / p >
< h4 id = "q-styletext_3" > Q: StyleText 合成数据效果不好？< a class = "headerlink" href = "#q-styletext_3" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ，
建议要合成的数据尺寸设置为32 x N。尺寸相差不多的数据也可以生成， < / p >
< h3 id = "25" > 2.5 预训练模型与微调< a class = "headerlink" href = "#25" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "qbackbone" > Q： ？ < a class = "headerlink" href = "#qbackbone" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， ， ， ， ， ， < / p >
< p > （ ） ， ， ， ， ， < / p >
< p > （ ） ， ， ， < / p >
< h4 id = "q_25" > Q: 参照文档做实际项目时，是重新训练还是在官方训练的基础上进行训练？具体如何操作？< a class = "headerlink" href = "#q_25" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ： ， < / p >
< h4 id = "q-inferencepdiparams-inferencepdmodel" > Q: 下载的识别模型解压后缺失文件， < a class = "headerlink" href = "#q-inferencepdiparams-inferencepdmodel" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "q-checkpointsload" > Q: 为什么在checkpoints中load下载的预训练模型会报错？ < a class = "headerlink" href = "#q-checkpointsload" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：这里有两个不同的概念：< / p >
< p > pretrained_model： ， ， ，
checkpoints： ， ， ， ， < / p >
< h4 id = "q-finetune" > Q: 如何对检测模型finetune， ？ < a class = "headerlink" href = "#q-finetune" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ： < a href = "https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/faq/train_cn.html#id4" > https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/faq/train_cn.html#id4< / a > < / p >
< p > 如果对某些层使用更小的学习率学习，静态图里还不是很方便，一个方法是在参数初始化的时候，给权重的属性设置固定的学习率，参考：< a href = "https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/fluid/param_attr/ParamAttr_cn.html#paramattr" > https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/api/paddle/fluid/param_attr/ParamAttr_cn.html#paramattr< / a > < / p >
< p > 实际上我们实验发现， ， ， < / p >
< h3 id = "26" > 2.6 模型超参调整< a class = "headerlink" href = "#26" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-db640" > Q: DB检测训练输入尺寸640， ？ < a class = "headerlink" href = "#q-db640" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， < / p >
< h4 id = "q-32" > Q: 预处理部分， ？ < a class = "headerlink" href = "#q-32" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， < / p >
< h4 id = "q-stride2-1" > Q: 在识别模型中， ？ < a class = "headerlink" href = "#q-stride2-1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : stride为(2, 1)， （ ） ， （ ） ， ， ， < / p >
< h4 id = "qshape" > Q： ， ？ < a class = "headerlink" href = "#qshape" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ： < / p >
< p > （ ） < / p >
< p > （ ） ， ： ， < / p >
< h4 id = "q_26" > Q： ， ， < a class = "headerlink" href = "#q_26" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：可以在命令中加入 --det_db_unclip_ratio ， ， ， ， < / p >
< h3 id = "27" > 2.7 模型结构< a class = "headerlink" href = "#27" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "qlstm" > Q： ？ < a class = "headerlink" href = "#qlstm" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， < / p >
< h4 id = "qlstmgru" > Q： ？ < a class = "headerlink" href = "#qlstmgru" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "qcrnnbackbonedensenetresnet_vd" > Q： ， ， ？ < a class = "headerlink" href = "#qcrnnbackbonedensenetresnet_vd" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， （ ） （ ） ， ， ， ， ， ， < / p >
< h4 id = "q-backbone" > Q: 如何根据不同的硬件平台选用不同的backbone？ < a class = "headerlink" href = "#q-backbone" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleClas/tree/release/2.0/docs/zh_CN/models" > PaddleClas模型速度-精度图< / a > 。< / p >
< h3 id = "28-pp-ocr" > 2.8 PP-OCR系统< a class = "headerlink" href = "#28-pp-ocr" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-pp-ocrse" > Q: 在PP-OCR系统中， ？ < a class = "headerlink" href = "#q-pp-ocrse" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， ， ， ， ， ， ， ， ， ， < a href = "https://arxiv.org/abs/2009.09941" > https://arxiv.org/abs/2009.09941< / a > .< / p >
< h4 id = "q-pp-ocr" > Q: PP-OCR系统中， ？ < a class = "headerlink" href = "#q-pp-ocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：文本检测的结果有置信度，由于推理过程中没有使用，所以没有显示的返回到最终结果中。如果需要文本检测结果的置信度，可以在< a href = "../../ppocr/postprocess/db_postprocess.py" > 文本检测DB的后处理代码< / a > 的155行， ， < a href = "../../tools/infer/predict_det.py" > 检测预测代码< / a > 的197行， < / p >
< h4 id = "q-db" > Q: DB文本检测， ？ < a class = "headerlink" href = "#q-db" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：特征提取网络金字塔构建的部分:< a href = "../../ppocr/modeling/necks/db_fpn.py" > 代码位置< / a > 。ppocr/modeling文件夹里面是组网相关的代码， ； ； ； ； < a href = "./tree.md" > 代码组织结构< / a > 。< / p >
< h4 id = "qpaddleocr" > Q： ？ < a class = "headerlink" href = "#qpaddleocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， < / p >
< h4 id = "q_27" > Q: 目前知识蒸馏有哪些主要的实践思路？< a class = "headerlink" href = "#q_27" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ： < / p >
< ol >
< li > 基于输出结果的蒸馏， （ ） （ ） < / li >
< li > 基于特征图的蒸馏，即让学生模型学习教师模型中间层的特征图，拟合中间层的一些特征。< / li >
< li > 基于关系的蒸馏， （ ） ， ， ， ， < / li >
< / ol >
< p > 当然，知识蒸馏方法日新月异，也欢迎大家提出更多的总结与建议。< / p >
< h4 id = "q-preds_idx-predsargmaxaxis2beam-search" > Q: 文字识别模型模型的输出矩阵需要进行解码才能得到识别的文本。代码中实现为preds_idx = preds.argmax(axis=2)， ， ， ？ < a class = "headerlink" href = "#q-preds_idx-predsargmaxaxis2beam-search" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， < / p >
< h3 id = "29" > 2.9 端到端< a class = "headerlink" href = "#29" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-pgnet" > Q: 端到端算法PGNet是否支持中文识别， ？ < a class = "headerlink" href = "#q-pgnet" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， （ ） ， ， ， ， < / p >
< h4 id = "q-pgnet_1" > Q: 端到端算法PGNet提供了两种后处理方式， ？ < a class = "headerlink" href = "#q-pgnet_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 两种后处理的区别主要在于速度的推理， ， ， ， ， < / p >
< h4 id = "q-pgneteval" > Q: 使用PGNet进行eval报错？ < a class = "headerlink" href = "#q-pgneteval" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 需要注意， ， ， ： < / p >
< ul >
< li > A模式： ， < / li >
< li > B模式： ， < / li >
< / ul >
< h4 id = "q-pgnet_2" > Q: PGNet有中文预训练模型吗？ < a class = "headerlink" href = "#q-pgnet_2" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 目前我们尚未提供针对中文的预训练模型，如有需要，可以尝试自己训练。具体需要修改的地方有：< / p >
< ol >
< li > < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/configs/e2e/e2e_r50_vd_pg.yml#L23-L24" > config文件中< / a > ，字典文件路径及语种设置；< / li >
< li > < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/ppocr/modeling/heads/e2e_pg_head.py#L181" > 网络结构中< / a > ， < code > out_channels< / code > 修改为字典中的字符数目+1（ ） ； < / li >
< li > < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/ppocr/losses/e2e_pg_loss.py#L93" > loss中< / a > ，修改< code > 37< / code > 为字典中的字符数目+1（ ） ； < / li >
< / ol >
< h4 id = "q-pgnet_3" > Q: 用于PGNet的训练集， ？ < a class = "headerlink" href = "#q-pgnet_3" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : PGNet支持多点标注， ， （ ） ， ， ，
我们提供的， < / p >
< h4 id = "q-pgnet_4" > Q: 用PGNet做进行端到端训练时， < a class = "headerlink" href = "#q-pgnet_4" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 目前代码要求标注为统一的点数。< / p >
< h3 id = "210" > 2.10 模型效果与效果不一致< a class = "headerlink" href = "#210" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-pp-ocr_1" > Q: PP-OCR检测效果不好， ？ < a class = "headerlink" href = "#q-pp-ocr_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：
如果在你的场景上检测效果不可用， ；
如果图像过大， ， ， ， ；
检测框大小过于紧贴文字或检测框过大， ， ， ；
检测框存在很多漏检问题， ， ，
其他方法可以选择use_dilation为True， ， ， ； < / p >
< h4 id = "q2126" > Q： ， < a class = "headerlink" href = "#q2126" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：可以主要参考可视化效果，通用模型更倾向于检测一整行文字，轻量级可能会有一行文字被分成两段检测的情况，不是数量越多，效果就越好。< / p >
< h4 id = "q-db_1" > Q: DB有些框太贴文本了反而去掉了一些文本的边角影响识别， ？ < a class = "headerlink" href = "#q-db_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< h4 id = "q-infer" > Q: 使用合成数据精调小模型后， ， ， ？ < a class = "headerlink" href = "#q-infer" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ； < / p >
< p > （ ） ， ， ， ， ， ； < / p >
< p > （ ） ， ， ， < / p >
< h4 id = "q_28" > Q: 表格识别中，如何提高单字的识别结果？< a class = "headerlink" href = "#q_28" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 首先需要确认一下检测模型有没有有效的检测出单个字符，如果没有的话，需要在训练集当中添加相应的单字数据集。< / p >
< h4 id = "q-dygraphrelease20" > Q: 动态图分支(dygraph,release/2.0)，训练模型和推理模型效果不一致< a class = "headerlink" href = "#q-dygraphrelease20" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ： ， ， ； ： < / p >
< ol >
< li > 预处理函数设置的不一致< / li >
< li > 后处理参数不一致 repo中config.yml文件的前后处理参数和inference预测默认的超参数有不一致的地方， ， < / li >
< / ol >
< h4 id = "q-detinferenceeval" > Q: 自己训练的det模型， ， ， ？ < a class = "headerlink" href = "#q-detinferenceeval" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， （ ， ， ， < / p >
< h4 id = "q_29" > Q: 训练模型和测试模型的检测结果差距较大< a class = "headerlink" href = "#q_29" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "q-paddleocrpythonc" > Q: PaddleOCR模型Python端预测和C++预测结果不一致？< a class = "headerlink" href = "#q-paddleocrpythonc" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， < / p >
< p > 用户总结的排查步骤：< a href = "https://github.com/PaddlePaddle/PaddleOCR/issues/2470" > https://github.com/PaddlePaddle/PaddleOCR/issues/2470< / a > < / p >
< h3 id = "211" > 2.11 训练调试与配置文件< a class = "headerlink" href = "#211" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-epoch" > Q: 某个类别的样本比较少， ， ， ？ < a class = "headerlink" href = "#q-epoch" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< h4 id = "q_30" > Q： ， < a class = "headerlink" href = "#q_30" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：数据格式有问题，”###” 表示要被忽略的文本区域，所以你的数据都被跳过了，可以换成其他任意字符或者就写个空的。< / p >
< h4 id = "q_31" > Q： ？ < a class = "headerlink" href = "#q_31" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< h4 id = "q_32" > Q： ， < a class = "headerlink" href = "#q_32" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， < / p >
< h4 id = "q-config-ymlratio_list" > Q: config yml文件中的ratio_list参数的作用是什么？ < a class = "headerlink" href = "#q-config-ymlratio_list" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 在动态图中， ， ， ， ， < / p >
< p > 静态图检测数据采样的逻辑与动态图不同，但基本不影响训练精度。< / p >
< p > 在静态图中，使用 检测 dataloader读取数据时， ， ， ， ， ， ， < / p >
< h4 id = "q-iaa" > Q: iaa里面添加的数据增强方式， ？ ？ < a class = "headerlink" href = "#q-iaa" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ： < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/0ccc1720c252beb277b9e522a1b228eb6abffb8a/configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml#L82" > 这里< / a > 。其中{ 'type': Fliplr, 'args': { 'p': 0.5 } } p是概率。新增数据增强， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release%2F2.1/doc/doc_ch/add_new_algorithm.md#%E6%95%B0%E6%8D%AE%E5%8A%A0%E8%BD%BD%E5%92%8C%E5%A4%84%E7%90%86" > 这个方法< / a > < / p >
< h4 id = "q_33" > Q: 怎么加速训练过程呢？< a class = "headerlink" href = "#q_33" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/doc/doc_ch/distributed_training.md" > 分布式训练教程文档< / a > 。< / p >
< h4 id = "q-finetune_1" > Q: 一些特殊场景的数据识别效果差， ， ？ < a class = "headerlink" href = "#q-finetune_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：您可以合成一些接近使用场景的数据用于训练。
我们计划推出基于特定场景的文本数据合成工具， < / p >
< h4 id = "q-paddleocr_2" > Q: PaddleOCR可以识别灰度图吗？ < a class = "headerlink" href = "#q-paddleocr_2" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ，
或者将单通道图像转换为三通道图像再识别。例如， < / p >
< h4 id = "q_34" > Q: 如何合成手写中文数据集？< a class = "headerlink" href = "#q_34" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 手写数据集可以通过手写单字数据集合成得到。随机选取一定数量的单字图片和对应的label， ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/a72d6f23be9979e0c103d911a9dca3e4613e8ccf/doc/doc_ch/handwritten_datasets.md" > 手写数据集< / a > 。< / p >
< h4 id = "qpaddleocr200step" > Q： ？ < a class = "headerlink" href = "#qpaddleocr200step" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "q-paddleocrcosine_decay" > Q: PaddleOCR在训练的时候一直使用cosine_decay的学习率下降策略， ？ < a class = "headerlink" href = "#q-paddleocrcosine_decay" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， < / p >
< h4 id = "q-cosine" > Q: Cosine学习率的更新策略是怎样的？ ？ < a class = "headerlink" href = "#q-cosine" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : Cosine学习率的说明可以参考< a href = "https://www.paddlepaddle.org.cn/documentation/docs/zh/api/paddle/optimizer/lr/CosineAnnealingDecay_cn.html#cosineannealingdecay" > 这里< / a > < / p >
< p > 在PaddleOCR中， ，
学习率的更新会和总的iter数量有关。当iter比较大时， < / p >
< h4 id = "q-cosinewarmup" > Q: 之前的CosineWarmup方法为什么不见了？ < a class = "headerlink" href = "#q-cosinewarmup" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 我们对代码结构进行了调整， ，
例如下面的代码， ： < / p >
< div class = "language-text highlight" > < pre > < span > < / span > < code > < span id = "__span-0-1" > < a id = "__codelineno-0-1" name = "__codelineno-0-1" href = "#__codelineno-0-1" > < / a > lr:
< / span > < span id = "__span-0-2" > < a id = "__codelineno-0-2" name = "__codelineno-0-2" href = "#__codelineno-0-2" > < / a > name: Cosine
< / span > < span id = "__span-0-3" > < a id = "__codelineno-0-3" name = "__codelineno-0-3" href = "#__codelineno-0-3" > < / a > learning_rate: 0.001
< / span > < span id = "__span-0-4" > < a id = "__codelineno-0-4" name = "__codelineno-0-4" href = "#__codelineno-0-4" > < / a > warmup_epoch: 2
< / span > < / code > < / pre > < / div >
< h4 id = "q-warmup" > Q: 训练识别和检测时学习率要加上warmup， ？ < a class = "headerlink" href = "#q-warmup" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : Warmup机制先使学习率从一个较小的值逐步升到一个较大的值， ， ， < / p >
< h4 id = "q-dygraph" > Q: 关于dygraph分支中， ， ？ < a class = "headerlink" href = "#q-dygraph" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：可以参考< a href = "../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml" > 配置文件< / a > 在< code > Train['dataset']['transforms']< / code > 添加RecAug字段， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/issues/1744" > ISSUE 1744< / a > 。< / p >
< h4 id = "q_35" > Q: 训练过程中，训练程序意外退出/挂起，应该如何解决？< a class = "headerlink" href = "#q_35" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 考虑内存， （ ） ， ， ， ， ， < / p >
< h4 id = "q-log" > Q: 训练程序启动后直到结束， ？ < a class = "headerlink" href = "#q-log" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 可以从以下三方面考虑：
1. 检查训练进程是否正常退出、显存占用是否释放、是否有残留进程， ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/doc/doc_ch/installation.md" > 安装< / a > 。
2. 检查数据集的数据量是否太小， ， ， ，
3. 如果使用私有数据集训练， < / p >
< h4 id = "q-num-workers" > Q: 配置文件中的参数num workers是什么意思， ？ < a class = "headerlink" href = "#q-num-workers" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 训练数据的读取需要硬盘IO， ， ， ， ， ， ， ， ， ， ， ， ， ， ， < / p >
< h3 id = "212" > 2.12 预测< a class = "headerlink" href = "#212" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "q-paddleocrtest_batch_size_per_card1" > Q: 为什么PaddleOCR检测预测是只支持一张图片测试？ < a class = "headerlink" href = "#q-paddleocrtest_batch_size_per_card1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， < / p >
< h4 id = "q-paddleocrtensorrt" > Q: PaddleOCR支持tensorrt推理吗？ < a class = "headerlink" href = "#q-paddleocrtensorrt" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "q-tensorrtpaddleocr" > Q: 如何使用TensorRT加速PaddleOCR预测？ < a class = "headerlink" href = "#q-tensorrtpaddleocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/3ec57e8df9263de6fa897e33d2d91bc5d0849ef3/tools/infer/utility.py#L37" > --use_tensorrt=True< / a > 即可，
C++TensorRT预测需要使用支持TRT的预测库并在编译时打开< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/3ec57e8df9263de6fa897e33d2d91bc5d0849ef3/deploy/cpp_infer/tools/build.sh#L15" > -DWITH_TENSORRT=ON< / a > 。
如果想修改其他分支代码支持TensorRT预测， < a href = "https://github.com/PaddlePaddle/PaddleOCR/pull/2921" > PR< / a > 。< / p >
< p > 注： < / p >
< h4 id = "q_36" > Q: 为什么识别模型做预测的时候，预测图片的数量数量还会影响预测的精度< a class = "headerlink" href = "#q_36" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ： < / p >
< div class = "language-bash highlight" > < table class = "highlighttable" > < tr > < td class = "linenos" > < div class = "linenodiv" > < pre > < span > < / span > < span class = "normal" > < a href = "#__codelineno-1-1" > 1< / a > < / span > < / pre > < / div > < / td > < td class = "code" > < div > < pre > < span > < / span > < code > < span id = "__span-1-1" > < a id = "__codelineno-1-1" name = "__codelineno-1-1" > < / a > python3< span class = "w" > < / span > tools/infer/predict_rec.py< span class = "w" > < / span > --image_dir< span class = "o" > =< / span > < span class = "s2" > " ./doc/imgs_words/ch/word_4.jpg" < / span > < span class = "w" > < / span > --rec_model_dir< span class = "o" > =< / span > < span class = "s2" > " ./ch_PP-OCRv3_rec_infer/" < / span > < span class = "w" > < / span > --rec_batch_num< span class = "o" > =< / span > < span class = "m" > 1< / span >
< / span > < / code > < / pre > < / div > < / td > < / tr > < / table > < / div >
< h3 id = "213" > 2.13 推理部署< a class = "headerlink" href = "#213" title = "Permanent link" > ¶ < / a > < / h3 >
< h4 id = "qpaddleocr_1" > Q： ？ < a class = "headerlink" href = "#qpaddleocr_1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：目前推理方式支持基于训练引擎推理和基于预测引擎推理。< / p >
< p > （ ） ， ， ， < / p >
< p > （ ） ， ， ， < / p >
< h4 id = "qpaddleocrcputenorrtgpu" > Q： ， ， ？ ？ < a class = "headerlink" href = "#qpaddleocrcputenorrtgpu" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： （ ） ； ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/dygraph/tools/infer/utility.py#L99" > 参考代码< / a > ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/tree/dygraph/deploy/cpp_infer" > 文档< / a > < / p >
< p > （ ） ， < / p >
< h4 id = "qhubservingpdserving" > Q： ？ < a class = "headerlink" href = "#qhubservingpdserving" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， ， ， ， < / p >
< h4 id = "q-paddle-hub-serving-imgpathimgurl" > Q: 目前paddle hub serving 只支持 imgpath， < a class = "headerlink" href = "#q-paddle-hub-serving-imgpathimgurl" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：图片是在< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/67ef25d593c4eabfaaceb22daade4577f53bed81/deploy/hubserving/ocr_system/module.py#L55" > 这里< / a > 读取的， 可以参考下面的写法， < / p >
< div class = "language-text highlight" > < pre > < span > < / span > < code > < span id = "__span-2-1" > < a id = "__codelineno-2-1" name = "__codelineno-2-1" href = "#__codelineno-2-1" > < / a > response = request.urlopen(' http://i1.whymtj.com/uploads/tu/201902/9999/52491ae4ba.jpg' )
< / span > < span id = "__span-2-2" > < a id = "__codelineno-2-2" name = "__codelineno-2-2" href = "#__codelineno-2-2" > < / a > img_array = np.array(bytearray(response.read()), dtype=np.uint8)
< / span > < span id = "__span-2-3" > < a id = "__codelineno-2-3" name = "__codelineno-2-3" href = "#__codelineno-2-3" > < / a > img = cv.imdecode(img_array, -1)
< / span > < / code > < / pre > < / div >
< h4 id = "q-c-ocr" > Q: C++ 端侧部署可以只对OCR的检测部署吗？ < a class = "headerlink" href = "#q-c-ocr" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/de3e2e7cd3b8b65ee02d7a41e570fa5b511a3c1d/deploy/cpp_infer/src/main.cpp#L72" > 参考< / a > < / p >
< h4 id = "q_37" > Q： ， ？ < a class = "headerlink" href = "#q_37" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ，
ocr_det：
ocr_rec: 识别预测
ocr_system: 检测识别串联预测< / p >
< h4 id = "q-litenb" > Q: lite预测库和nb模型版本不匹配， ？ < a class = "headerlink" href = "#q-litenb" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release%2F2.1/deploy/lite/readme.md" > 移动端部署教程< / a > 。< / p >
< h4 id = "qpaddleocrsdk" > Q： < a class = "headerlink" href = "#qpaddleocrsdk" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "qpaddleocrtest_batch_size_per_card1" > Q： ？ < a class = "headerlink" href = "#qpaddleocrtest_batch_size_per_card1" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， < / p >
< h4 id = "q_38" > Q： ， ？ < a class = "headerlink" href = "#q_38" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ：第一张图需要显存资源初始化，耗时较多。完成模型加载后，之后的预测时间会明显缩短。< / p >
< h4 id = "q-paddle-lite" > Q: 采用Paddle-Lite进行端侧部署， ， < a class = "headerlink" href = "#q-paddle-lite" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， < / p >
< h4 id = "q-paddleocr_3" > Q: 如何多进程运行paddleocr？ < a class = "headerlink" href = "#q-paddleocr_3" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， ， ， ， ， < / p >
< h4 id = "q_39" > Q: 如何多进程预测？< a class = "headerlink" href = "#q_39" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 近期PaddleOCR新增了< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/a312647be716776c1aac33ff939ae358a39e8188/tools/infer/utility.py#L103" > 多进程预测控制参数< / a > ， < code > use_mp< / code > 表示是否使用多进程，< code > total_process_num< / code > 表示在使用多进程时的进程数。具体使用方式请参考< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/doc/doc_ch/inference.md#1-%E8%B6%85%E8%BD%BB%E9%87%8F%E4%B8%AD%E6%96%87ocr%E6%A8%A1%E5%9E%8B%E6%8E%A8%E7%90%86" > 文档< / a > 。< / p >
< h4 id = "q-paddleocrt4" > Q: 怎么解决paddleOCR在T4卡上有越预测越慢的情况？ < a class = "headerlink" href = "#q-paddleocrt4" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： < / p >
< ol >
< li > T4 GPU没有主动散热， ， < / li >
< li > T4在不使用的时候， ， ， < / li >
< / ol >
< div class = "language-text highlight" > < pre > < span > < / span > < code > < span id = "__span-3-1" > < a id = "__codelineno-3-1" name = "__codelineno-3-1" href = "#__codelineno-3-1" > < / a > nvidia-smi -i 0 -pm ENABLED
< / span > < span id = "__span-3-2" > < a id = "__codelineno-3-2" name = "__codelineno-3-2" href = "#__codelineno-3-2" > < / a > nvidia-smi --lock-gpu-clocks=1590 -i 0
< / span > < / code > < / pre > < / div >
< h4 id = "q-windowscpp-inferencepaddle_fluiddllopencv_world346dll" > Q: 在windows上进行cpp inference的部署时， < code > paddle_fluid.dll< / code > 和< code > opencv_world346.dll< / code > < a class = "headerlink" href = "#q-windowscpp-inferencepaddle_fluiddllopencv_world346dll" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ： < / p >
< ol >
< li > 将paddle预测库和opencv库的地址添加到系统环境变量中。< / li >
< li > 将提示缺失的dll文件拷贝到编译产出的< code > ocr_system.exe< / code > 文件夹中。< / li >
< / ol >
< h4 id = "q-winc" > Q: win下C++部署中文识别乱码的解决方法< a class = "headerlink" href = "#q-winc" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : win下编码格式不是utf8,而ppocr_keys_v1.txt的编码格式的utf8， < / p >
< h4 id = "q-windows-3060gpu" > Q: windows 3060显卡GPU模式启动 加载模型慢< a class = "headerlink" href = "#q-windows-3060gpu" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 30系列的显卡需要使用cuda11。< / p >
< h4 id = "qmac" > Q： ， ？ < a class = "headerlink" href = "#qmac" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ： < a href = "https://paddle-inference-lib.bj.bcebos.com/mac/2.0.0/cpu_avx_openblas/paddle_inference.tgz" > https://paddle-inference-lib.bj.bcebos.com/mac/2.0.0/cpu_avx_openblas/paddle_inference.tgz< / a > < / p >
< h4 id = "q_40" > Q： ？ < a class = "headerlink" href = "#q_40" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > ： ， < / p >
< h4 id = "q-hub_serving" > Q: 使用hub_serving部署， ， ？ < a class = "headerlink" href = "#q-hub_serving" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 首先， ， ； ， （ ） ， ， （ ） < / p >
< h4 id = "q-paddlelite" > Q: 在使用PaddleLite进行预测部署时， < a class = "headerlink" href = "#q-paddlelite" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 请检查模型转换时所用PaddleLite的版本， ， < / p >
< h4 id = "q_41" > Q: 预测时显存爆炸、内存泄漏问题？< a class = "headerlink" href = "#q_41" title = "Permanent link" > ¶ < / a > < / h4 >
< p > < strong > A< / strong > : 打开显存/内存优化开关< code > enable_memory_optim< / code > 可以解决该问题，相关代码已合入，< a href = "https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.1/tools/infer/utility.py#L153" > 查看详情< / a > 。< / p >
< aside class = "md-source-file" >
< span class = "md-source-file__fact" >
< span class = "md-icon" title = "最后更新" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M21 13.1c-.1 0-.3.1-.4.2l-1 1 2.1 2.1 1-1c.2-.2.2-.6 0-.8l-1.3-1.3c-.1-.1-.2-.2-.4-.2m-1.9 1.8-6.1 6V23h2.1l6.1-6.1zM12.5 7v5.2l4 2.4-1 1L11 13V7zM11 21.9c-5.1-.5-9-4.8-9-9.9C2 6.5 6.5 2 12 2c5.3 0 9.6 4.1 10 9.3-.3-.1-.6-.2-1-.2s-.7.1-1 .2C19.6 7.2 16.2 4 12 4c-4.4 0-8 3.6-8 8 0 4.1 3.1 7.5 7.1 7.9l-.1.2z" / > < / svg >
< / span >
2024-12-12 07:39:44 +00:00
< span class = "git-revision-date-localized-plugin git-revision-date-localized-plugin-date" > 2024年12月12日< / span >
2024-10-24 15:00:38 +00:00
< / span >
< span class = "md-source-file__fact" >
< span class = "md-icon" title = "创建日期" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M14.47 15.08 11 13V7h1.5v5.25l3.08 1.83c-.41.28-.79.62-1.11 1m-1.39 4.84c-.36.05-.71.08-1.08.08-4.42 0-8-3.58-8-8s3.58-8 8-8 8 3.58 8 8c0 .37-.03.72-.08 1.08.69.1 1.33.32 1.92.64.1-.56.16-1.13.16-1.72 0-5.5-4.5-10-10-10S2 6.5 2 12s4.47 10 10 10c.59 0 1.16-.06 1.72-.16-.32-.59-.54-1.23-.64-1.92M18 15v3h-3v2h3v3h2v-3h3v-2h-3v-3z" / > < / svg >
< / span >
2024-12-12 07:39:44 +00:00
< span class = "git-revision-date-localized-plugin git-revision-date-localized-plugin-date" > 2024年12月12日< / span >
2024-10-24 15:00:38 +00:00
< / span >
< span class = "md-source-file__fact" >
< span class = "md-icon" title = "贡献者" >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M12 2A10 10 0 0 0 2 12c0 4.42 2.87 8.17 6.84 9.5.5.08.66-.23.66-.5v-1.69c-2.77.6-3.36-1.34-3.36-1.34-.46-1.16-1.11-1.47-1.11-1.47-.91-.62.07-.6.07-.6 1 .07 1.53 1.03 1.53 1.03.87 1.52 2.34 1.07 2.91.83.09-.65.35-1.09.63-1.34-2.22-.25-4.55-1.11-4.55-4.92 0-1.11.38-2 1.03-2.71-.1-.25-.45-1.29.1-2.64 0 0 .84-.27 2.75 1.02.79-.22 1.65-.33 2.5-.33s1.71.11 2.5.33c1.91-1.29 2.75-1.02 2.75-1.02.55 1.35.2 2.39.1 2.64.65.71 1.03 1.6 1.03 2.71 0 3.82-2.34 4.66-4.57 4.91.36.31.69.92.69 1.85V21c0 .27.16.59.67.5C19.14 20.16 22 16.42 22 12A10 10 0 0 0 12 2" / > < / svg >
< / span >
< span > GitHub< / span >
< nav >
< a href = "https://github.com/GreatV" class = "md-author" title = "@GreatV" >
< img src = "https://avatars.githubusercontent.com/u/17264618?v=4&size=72" alt = "GreatV" >
< / a >
< a href = "https://github.com/web-flow" class = "md-author" title = "@web-flow" >
< img src = "https://avatars.githubusercontent.com/u/19864447?v=4&size=72" alt = "web-flow" >
< / a >
< a href = "https://github.com/SWHL" class = "md-author" title = "@SWHL" >
< img src = "https://avatars.githubusercontent.com/u/28639377?v=4&size=72" alt = "SWHL" >
< / a >
< / nav >
< / span >
< / aside >
< h2 id = "__comments" > 评论< / h2 >
<!-- Insert generated snippet here -->
< script src = "https://giscus.app/client.js" data-repo = "PaddlePaddle/PaddleOCR"
data-repo-id="MDEwOlJlcG9zaXRvcnkyNjIyOTYxMjI=" data-category="Q& A" data-category-id="DIC_kwDOD6JSOs4COrbO"
data-mapping="pathname" data-strict="0" data-reactions-enabled="1" data-emit-metadata="0" data-input-position="top"
data-theme="preferred_color_scheme" data-lang="en" data-loading="lazy" crossorigin="anonymous" async>
< / script >
<!-- Synchronize Giscus theme with palette -->
< script >
var giscus = document.querySelector("script[src*=giscus]")
// Set palette on initial load
var palette = __md_get("__palette")
if (palette & & typeof palette.color === "object") {
var theme = palette.color.scheme === "slate"
? "transparent_dark"
: "light"
// Instruct Giscus to set theme
giscus.setAttribute("data-theme", theme)
}
// Register event handlers after documented loaded
document.addEventListener("DOMContentLoaded", function () {
var ref = document.querySelector("[data-md-component=palette]")
ref.addEventListener("change", function () {
var palette = __md_get("__palette")
if (palette & & typeof palette.color === "object") {
var theme = palette.color.scheme === "slate"
? "transparent_dark"
: "light"
// Instruct Giscus to change theme
var frame = document.querySelector(".giscus-frame")
frame.contentWindow.postMessage(
{ giscus: { setConfig: { theme } } },
"https://giscus.app"
)
}
})
})
< / script >
< / article >
< / div >
< script > var tabs = _ _md _get ( "__tabs" ) ; if ( Array . isArray ( tabs ) ) e : for ( var set of document . querySelectorAll ( ".tabbed-set" ) ) { var labels = set . querySelector ( ".tabbed-labels" ) ; for ( var tab of tabs ) for ( var label of labels . getElementsByTagName ( "label" ) ) if ( label . innerText . trim ( ) === tab ) { var input = document . getElementById ( label . htmlFor ) ; input . checked = ! 0 ; continue e } } < / script >
< script > var target = document . getElementById ( location . hash . slice ( 1 ) ) ; target && target . name && ( target . checked = target . name . startsWith ( "__tabbed_" ) ) < / script >
< / div >
< button type = "button" class = "md-top md-icon" data-md-component = "top" hidden >
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 24 24" > < path d = "M13 20h-2V8l-5.5 5.5-1.42-1.42L12 4.16l7.92 7.92-1.42 1.42L13 8z" / > < / svg >
回到页面顶部
< / button >
< / main >
< footer class = "md-footer" >
< nav class = "md-footer__inner md-grid" aria-label = "页脚" >
< a href = "datasets/kie_datasets.html" class = "md-footer__link md-footer__link--prev" aria-label = "上一页: 关键信息提取数据集" >
< div class = "md-footer__button md-icon" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 320 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M41.4 233.4c-12.5 12.5-12.5 32.8 0 45.3l160 160c12.5 12.5 32.8 12.5 45.3 0s12.5-32.8 0-45.3L109.3 256l137.3-137.4c12.5-12.5 12.5-32.8 0-45.3s-32.8-12.5-45.3 0l-160 160z" / > < / svg >
2024-10-24 15:00:38 +00:00
< / div >
< div class = "md-footer__title" >
< span class = "md-footer__direction" >
上一页
< / span >
< div class = "md-ellipsis" >
关键信息提取数据集
< / div >
< / div >
< / a >
< a href = "community/community_contribution.html" class = "md-footer__link md-footer__link--next" aria-label = "下一页: 社区贡献" >
< div class = "md-footer__title" >
< span class = "md-footer__direction" >
下一页
< / span >
< div class = "md-ellipsis" >
社区贡献
< / div >
< / div >
< div class = "md-footer__button md-icon" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 320 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M278.6 233.4c12.5 12.5 12.5 32.8 0 45.3l-160 160c-12.5 12.5-32.8 12.5-45.3 0s-12.5-32.8 0-45.3L210.7 256 73.4 118.6c-12.5-12.5-12.5-32.8 0-45.3s32.8-12.5 45.3 0l160 160z" / > < / svg >
2024-10-24 15:00:38 +00:00
< / div >
< / a >
< / nav >
< div class = "md-footer-meta md-typeset" >
< div class = "md-footer-meta__inner md-grid" >
< div class = "md-copyright" >
< div class = "md-copyright__highlight" >
Copyright © 2024 Maintained by PaddleOCR PMC.
< / div >
Made with
< a href = "https://squidfunk.github.io/mkdocs-material/" target = "_blank" rel = "noopener" >
Material for MkDocs
< / a >
< / div >
< div class = "md-social" >
< a href = "https://github.com/PaddlePaddle/PaddleOCR" target = "_blank" rel = "noopener" title = "github.com" class = "md-social__link" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 496 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M165.9 397.4c0 2-2.3 3.6-5.2 3.6-3.3.3-5.6-1.3-5.6-3.6 0-2 2.3-3.6 5.2-3.6 3-.3 5.6 1.3 5.6 3.6m-31.1-4.5c-.7 2 1.3 4.3 4.3 4.9 2.6 1 5.6 0 6.2-2s-1.3-4.3-4.3-5.2c-2.6-.7-5.5.3-6.2 2.3m44.2-1.7c-2.9.7-4.9 2.6-4.6 4.9.3 2 2.9 3.3 5.9 2.6 2.9-.7 4.9-2.6 4.6-4.6-.3-1.9-3-3.2-5.9-2.9M244.8 8C106.1 8 0 113.3 0 252c0 110.9 69.8 205.8 169.5 239.2 12.8 2.3 17.3-5.6 17.3-12.1 0-6.2-.3-40.4-.3-61.4 0 0-70 15-84.7-29.8 0 0-11.4-29.1-27.8-36.6 0 0-22.9-15.7 1.6-15.4 0 0 24.9 2 38.6 25.8 21.9 38.6 58.6 27.5 72.9 20.9 2.3-16 8.8-27.1 16-33.7-55.9-6.2-112.3-14.3-112.3-110.5 0-27.5 7.6-41.3 23.6-58.9-2.6-6.5-11.1-33.3 2.6-67.9 20.9-6.5 69 27 69 27 20-5.6 41.5-8.5 62.8-8.5s42.8 2.9 62.8 8.5c0 0 48.1-33.6 69-27 13.7 34.7 5.2 61.4 2.6 67.9 16 17.7 25.8 31.5 25.8 58.9 0 96.5-58.9 104.2-114.8 110.5 9.2 7.9 17 22.9 17 46.4 0 33.7-.3 75.4-.3 83.6 0 6.5 4.6 14.4 17.3 12.1C428.2 457.8 496 362.9 496 252 496 113.3 383.5 8 244.8 8M97.2 352.9c-1.3 1-1 3.3.7 5.2 1.6 1.6 3.9 2.3 5.2 1 1.3-1 1-3.3-.7-5.2-1.6-1.6-3.9-2.3-5.2-1m-10.8-8.1c-.7 1.3.3 2.9 2.3 3.9 1.6 1 3.6.7 4.3-.7.7-1.3-.3-2.9-2.3-3.9-2-.6-3.6-.3-4.3.7m32.4 35.6c-1.6 1.3-1 4.3 1.3 6.2 2.3 2.3 5.2 2.6 6.5 1 1.3-1.3.7-4.3-1.3-6.2-2.2-2.3-5.2-2.6-6.5-1m-11.4-14.7c-1.6 1-1.6 3.6 0 5.9s4.3 3.3 5.6 2.3c1.6-1.3 1.6-3.9 0-6.2-1.4-2.3-4-3.3-5.6-2" / > < / svg >
2024-10-24 15:00:38 +00:00
< / a >
< a href = "https://pypi.org/project/paddleocr/" target = "_blank" rel = "noopener" title = "pypi.org" class = "md-social__link" >
2024-12-12 07:39:44 +00:00
< svg xmlns = "http://www.w3.org/2000/svg" viewBox = "0 0 448 512" > <!-- ! Font Awesome Free 6.7.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2024 Fonticons, Inc. --> < path d = "M439.8 200.5c-7.7-30.9-22.3-54.2-53.4-54.2h-40.1v47.4c0 36.8-31.2 67.8-66.8 67.8H172.7c-29.2 0-53.4 25-53.4 54.3v101.8c0 29 25.2 46 53.4 54.3 33.8 9.9 66.3 11.7 106.8 0 26.9-7.8 53.4-23.5 53.4-54.3v-40.7H226.2v-13.6h160.2c31.1 0 42.6-21.7 53.4-54.2 11.2-33.5 10.7-65.7 0-108.6M286.2 404c11.1 0 20.1 9.1 20.1 20.3 0 11.3-9 20.4-20.1 20.4-11 0-20.1-9.2-20.1-20.4.1-11.3 9.1-20.3 20.1-20.3M167.8 248.1h106.8c29.7 0 53.4-24.5 53.4-54.3V91.9c0-29-24.4-50.7-53.4-55.6-35.8-5.9-74.7-5.6-106.8.1-45.2 8-53.4 24.7-53.4 55.6v40.7h106.9v13.6h-147c-31.1 0-58.3 18.7-66.8 54.2-9.8 40.7-10.2 66.1 0 108.6 7.6 31.6 25.7 54.2 56.8 54.2H101v-48.8c0-35.3 30.5-66.4 66.8-66.4m-6.7-142.6c-11.1 0-20.1-9.1-20.1-20.3.1-11.3 9-20.4 20.1-20.4 11 0 20.1 9.2 20.1 20.4s-9 20.3-20.1 20.3" / > < / svg >
2024-10-24 15:00:38 +00:00
< / a >
< / div >
< / div >
< / div >
< / footer >
< / div >
< div class = "md-dialog" data-md-component = "dialog" >
< div class = "md-dialog__inner md-typeset" > < / div >
< / div >
< script id = "__config" type = "application/json" > { "base" : "." , "features" : [ "announce.dismiss" , "content.tooltips" , "content.code.copy" , "content.tabs.link" , "content.footnote.tooltips" , "content.action.edit" , "content.action.view" , "navigation.expand" , "navigation.tabs" , "navigation.tabs.sticky" , "navigation.top" , "navigation.tracking" , "navigation.footer" , "navigation.indexes" , "search.highlight" , "search.share" , "search.suggest" , "toc.follow" ] , "search" : "assets/javascripts/workers/search.6ce7567c.min.js" , "translations" : { "clipboard.copied" : "\u5df2\u590d\u5236" , "clipboard.copy" : "\u590d\u5236" , "search.result.more.one" : "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c" , "search.result.more.other" : "\u5728\u8be5\u9875\u4e0a\u8fd8\u6709 # \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c" , "search.result.none" : "\u6ca1\u6709\u627e\u5230\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c" , "search.result.one" : "\u627e\u5230 1 \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c" , "search.result.other" : "# \u4e2a\u7b26\u5408\u6761\u4ef6\u7684\u7ed3\u679c" , "search.result.placeholder" : "\u952e\u5165\u4ee5\u5f00\u59cb\u641c\u7d22" , "search.result.term.missing" : "\u7f3a\u5c11" , "select.version" : "\u9009\u62e9\u5f53\u524d\u7248\u672c" } , "version" : { "provider" : "mike" } } < / script >
< script src = "assets/javascripts/bundle.83f73b43.min.js" > < / script >
< script src = "javascripts/katex.min.js" > < / script >
< script src = "https://unpkg.com/katex@0/dist/katex.min.js" > < / script >
< script src = "https://unpkg.com/katex@0/dist/contrib/auto-render.min.js" > < / script >
< / body >
< / html >
