History

Hongbin Sun f504637971 version1.0 (#57 ) * add sar, seg and other components * [feature]: add textsnake_drrg * documentation and dbnet related code * [feature]: add code for kie and textsnake config * [feature]: add CRNN and RobustScanner * Revert "documentation and dbnet related code" * [feature]: add textdet * [feature]: dbnet and docs * fix #9: [feature]: setting norms for contributing (#10) * fix #9: [feature]: setting norms for contributing * fix #9: [feature]: setting norms for contributing * fix #9: [feature]: setting norms for contributing * fix #9: [feature]: setting norms for contributing * fix #11: update docs (#12) * fix #11: update docs * fix #11: update datasets.md for kie * fix #13: update docs with toc * fix #13: link pr to issue * fix #13: rename section title * fix #13: rename section title (#16) * fix #17: update ckpt path of psenet (#18) * Enhance/synthtext pretrain (#20) * fix 19: add synthtext pretrained model * fix 19: setup.cfg linting * Format readme (#23) * Format readme Signed-off-by: lizz <lizz@sensetime.com> * try Signed-off-by: lizz <lizz@sensetime.com> * Remove redudant config link Signed-off-by: lizz <lizz@sensetime.com> * fix #21: refactor kie dataset & add show_results * fix #21: update sdmgr readme and config * fix #21: update readme of segocr * f-str Signed-off-by: lizz <lizz@sensetime.com> * format again Signed-off-by: lizz <lizz@sensetime.com> * Mkae sort_vertex public api Signed-off-by: lizz <lizz@sensetime.com> * fix #24: rm img_meta from inference (#25) * Fix typos (#26) * Fix typos Signed-off-by: lizz <lizz@sensetime.com> * Ohh Signed-off-by: lizz <lizz@sensetime.com> * [feature]: add nrtr (#28) * [feature]: add nrtr * Rename nrtr_top_dataset.py to nrtr_toy_dataset.py Co-authored-by: Hongbin Sun <hongbin306@gmail.com> * fix #29: update logo (#30) * Feature/iss 33 (#34) * fix #33: update dataset.md * fix #33: pytest for transformer related * Add Github CI Signed-off-by: lizz <lizz@sensetime.com> * rm old ci Signed-off-by: lizz <lizz@sensetime.com> * add contributing and code of conduct Signed-off-by: lizz <lizz@sensetime.com> * Fix ci Signed-off-by: lizz <lizz@sensetime.com> * fix Signed-off-by: lizz <lizz@sensetime.com> * fix Signed-off-by: lizz <lizz@sensetime.com> * Re-enable skipped test Signed-off-by: lizz <lizz@sensetime.com> * good contributing link Signed-off-by: lizz <lizz@sensetime.com> * Remove pytorch 1.3 Signed-off-by: lizz <lizz@sensetime.com> * Remove test dependency on tools Signed-off-by: lizz <lizz@sensetime.com> * fix #31: pytest pass * skip cuda Signed-off-by: lizz <lizz@sensetime.com> * try Signed-off-by: lizz <lizz@sensetime.com> * format Signed-off-by: lizz <lizz@sensetime.com> * again Signed-off-by: lizz <lizz@sensetime.com> * Revert "Remove pytorch 1.3" This reverts commit b8d65afea82a9ba9a5ee3315aa6816d21c137c91. * Revert me when rroi is moved to mmcv Signed-off-by: lizz <lizz@sensetime.com> * Revert "Revert "Remove pytorch 1.3"" This reverts commit 1629a64b9e5aecc5536698d988e7151e04c4772d. * Let it pass * fix #35: add nrtr readme; update nrtr config (#36) * fix #37: remove useless code (#38) * np.int -> np.int32 Signed-off-by: lizz <lizz@sensetime.com> * out_size -> output_size Signed-off-by: lizz <lizz@sensetime.com> * Add textdet unit tests (#43) * Fix #41: test fpn_cat * Fix #41: test fpn_cat * Fix #41: test fpn_cat * fix #40: add unit test for recog config, transforms, etc. (#44) * fix #45: remove useless (#46) * fix #47: add unit test for api (#48) * add Dockerfile (#50) * Textsnake tests (#51) * add textsnake unit tests * Remove usage of \ (#49) * Remove usage of \ Signed-off-by: lizz <lizz@sensetime.com> * rebase Signed-off-by: lizz <lizz@sensetime.com> * typos Signed-off-by: lizz <lizz@sensetime.com> * Remove test dependency on tools/ Signed-off-by: lizz <lizz@sensetime.com> * Remove usage of \ Signed-off-by: lizz <lizz@sensetime.com> * rebase Signed-off-by: lizz <lizz@sensetime.com> * typos Signed-off-by: lizz <lizz@sensetime.com> * Remove test dependency on tools/ Signed-off-by: lizz <lizz@sensetime.com> * typo Signed-off-by: lizz <lizz@sensetime.com> * KIE in keywords Signed-off-by: lizz <lizz@sensetime.com> * some renames Signed-off-by: lizz <lizz@sensetime.com> * kill isort skip Signed-off-by: lizz <lizz@sensetime.com> * aggregation discrimination Signed-off-by: lizz <lizz@sensetime.com> * aggregation discrimination Signed-off-by: lizz <lizz@sensetime.com> * tiny Signed-off-by: lizz <lizz@sensetime.com> * fix bug: model infer on cpu Co-authored-by: Hongbin Sun <hongbin306@gmail.com> * fix #52: update readme (#53) * fix #39: update crnn & robustscanner. (#54) * fix #55: update nrtr readme (#56) Co-authored-by: HolyCrap96 <theochan666@gmail.com> Co-authored-by: quincylin1 <quincylin.333@gmail.com> Co-authored-by: YueXy <yuexiaoyu@sensetime.com> Co-authored-by: yuexy <yuexy@users.noreply.github.com> Co-authored-by: jeffreykuang <kuangzhh@gmail.com> Co-authored-by: lizz <innerlee@users.noreply.github.com> Co-authored-by: lizz <lizz@sensetime.com> Co-authored-by: Theo Chan <46100303+HolyCrap96@users.noreply.github.com>		2021-04-06 14:56:33 +00:00
..
README.md	version1.0 (#57 )	2021-04-06 14:56:33 +00:00
nrtr_modality_toy.py	version1.0 (#57 )	2021-04-06 14:56:33 +00:00
nrtr_r31_academic.py	version1.0 (#57 )	2021-04-06 14:56:33 +00:00

README.md

NRTR

Introduction

[ALGORITHM]

@inproceedings{sheng2019nrtr,
  title={NRTR: A no-recurrence sequence-to-sequence model for scene text recognition},
  author={Sheng, Fenfen and Chen, Zhineng and Xu, Bo},
  booktitle={2019 International Conference on Document Analysis and Recognition (ICDAR)},
  pages={781--786},
  year={2019},
  organization={IEEE}
}

[BACKBONE]

@inproceedings{li2019show,
  title={Show, attend and read: A simple and strong baseline for irregular text recognition},
  author={Li, Hui and Wang, Peng and Shen, Chunhua and Zhang, Guyu},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={33},
  number={01},
  pages={8610--8617},
  year={2019}
}

Dataset

Train Dataset

trainset	instance_num	repeat_num	source
SynthText	7266686	1	synth
Syn90k	8919273	1	synth

Test Dataset

testset	instance_num	type
IIIT5K	3000	regular
SVT	647	regular
IC13	1015	regular
IC15	2077	irregular
SVTP	645	irregular
CT80	288	irregular

Results and Models

Methods	Backbone		Regular Text				Irregular Text		download
		IIIT5K	SVT	IC13		IC15	SVTP	CT80
NRTR	R31-1/16-1/8	93.9	90.0	93.5		74.5	78.5	86.5	model \| log

Notes:

R31-1/16-1/8 means the height of feature from backbone is 1/16 of input image, where 1/8 for width.