From b4678eb6571a56fbe5420fe676a5295aeda79d2c Mon Sep 17 00:00:00 2001
From: Xinyu Wang <45810070+xinke-wang@users.noreply.github.com>
Date: Thu, 5 May 2022 14:09:05 +0800
Subject: [PATCH] [Fix] Fix Data Converter Issues (#955)

* fix naf mask issue; fix lv path issue

* fix path

* fix ic13, ic11 path issue; fix cocotextv2 mask issue

* fix funsd format
---
 tools/data/textdet/cocotext_converter.py | 2 ++
 tools/data/textdet/lv_converter.py       | 3 ++-
 tools/data/textdet/naf_converter.py      | 5 +++--
 tools/data/textrecog/funsd_converter.py  | 4 ++--
 tools/data/textrecog/ic11_converter.py   | 4 ++--
 tools/data/textrecog/ic13_converter.py   | 4 ++--
 6 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/tools/data/textdet/cocotext_converter.py b/tools/data/textdet/cocotext_converter.py
index 049efd38..4c28b7ba 100644
--- a/tools/data/textdet/cocotext_converter.py
+++ b/tools/data/textdet/cocotext_converter.py
@@ -92,6 +92,8 @@ def collect_cocotext_info(root_path, split, print_every=1000):
                 w, h = math.ceil(w), math.ceil(h)
                 bbox = [x, y, w, h]
                 segmentation = [max(0, int(x)) for x in ann['mask']]
+                if len(segmentation) < 8 or len(segmentation) % 2 != 0:
+                    segmentation = [x, y, x + w, y, x + w, y + h, x, y + h]
                 anno = dict(
                     iscrowd=iscrowd,
                     category_id=1,
diff --git a/tools/data/textdet/lv_converter.py b/tools/data/textdet/lv_converter.py
index e14026ee..1d5dfa75 100644
--- a/tools/data/textdet/lv_converter.py
+++ b/tools/data/textdet/lv_converter.py
@@ -83,9 +83,10 @@ def load_img_info(files):
         '.')[0]
     # read imgs while ignoring orientations
     img = mmcv.imread(img_file, 'unchanged')
+    img_file = img_file.split('data/lv/')[1]
 
     img_info = dict(
-        file_name=osp.join(osp.basename(img_file)),
+        file_name=img_file,
         height=img.shape[0],
         width=img.shape[1],
         segm_file=osp.join(osp.basename(gt_file)))
diff --git a/tools/data/textdet/naf_converter.py b/tools/data/textdet/naf_converter.py
index cb800457..219950f7 100644
--- a/tools/data/textdet/naf_converter.py
+++ b/tools/data/textdet/naf_converter.py
@@ -139,14 +139,15 @@ def load_json_info(gt_file, img_info):
                 if anno['type'] == 'blank':
                     continue
 
-            xs, ys = [], []
+            xs, ys, segmentation = [], [], []
             for p in anno['poly_points']:
                 xs.append(p[0])
                 ys.append(p[1])
+                segmentation.append(p[0])
+                segmentation.append(p[1])
             x, y = max(0, min(xs)), max(0, min(ys))
             w, h = max(xs) - x, max(ys) - y
             bbox = [x, y, w, h]
-            segmentation = anno['poly_points']
 
             anno = dict(
                 iscrowd=0,
diff --git a/tools/data/textrecog/funsd_converter.py b/tools/data/textrecog/funsd_converter.py
index 409f0702..baa73e16 100644
--- a/tools/data/textrecog/funsd_converter.py
+++ b/tools/data/textrecog/funsd_converter.py
@@ -140,9 +140,9 @@ def generate_ann(root_path, split, image_infos, preserve_vertical, format):
 
     dst_image_root = osp.join(root_path, 'dst_imgs', split)
     if split == 'training':
-        dst_label_file = osp.join(root_path, 'train_label.txt')
+        dst_label_file = osp.join(root_path, f'train_label.{format}')
     elif split == 'test':
-        dst_label_file = osp.join(root_path, 'test_label.txt')
+        dst_label_file = osp.join(root_path, f'test_label.{format}')
     os.makedirs(dst_image_root, exist_ok=True)
 
     lines = []
diff --git a/tools/data/textrecog/ic11_converter.py b/tools/data/textrecog/ic11_converter.py
index e46b395a..bf638c72 100644
--- a/tools/data/textrecog/ic11_converter.py
+++ b/tools/data/textrecog/ic11_converter.py
@@ -39,7 +39,7 @@ def convert_annotations(root_path, split, format):
             'r',
             encoding='"utf-8-sig') as f:
         annos = f.readlines()
-    dst_image_root = osp.join(root_path, split)
+    dst_image_root = osp.join(root_path, split.lower())
     for anno in annos:
         # text may contain comma ','
         dst_img_name, word = anno.split(', "')
@@ -58,7 +58,7 @@ def convert_annotations(root_path, split, format):
         else:
             raise NotImplementedError
 
-    list_to_file(osp.join(root_path, f'{split}_label.{format}'), lines)
+    list_to_file(osp.join(root_path, f'{split.lower()}_label.{format}'), lines)
 
 
 def parse_args():
diff --git a/tools/data/textrecog/ic13_converter.py b/tools/data/textrecog/ic13_converter.py
index b6f0c57d..28ddc12f 100644
--- a/tools/data/textrecog/ic13_converter.py
+++ b/tools/data/textrecog/ic13_converter.py
@@ -39,7 +39,7 @@ def convert_annotations(root_path, split, format):
             'r',
             encoding='"utf-8-sig') as f:
         annos = f.readlines()
-    dst_image_root = osp.join(root_path, split)
+    dst_image_root = osp.join(root_path, split.lower())
     for anno in annos:
         # text may contain comma ','
         dst_img_name, word = anno.split(', "')
@@ -58,7 +58,7 @@ def convert_annotations(root_path, split, format):
         else:
             raise NotImplementedError
 
-    list_to_file(osp.join(root_path, f'{split}_label.{format}'), lines)
+    list_to_file(osp.join(root_path, f'{split.lower()}_label.{format}'), lines)
 
 
 def parse_args():