[Feature] MLT Dataset Preparer

dataset_zoo/mlt/metafile.yml

@@ -0,0 +1,39 @@
+Name: 'MLT 2017 (ICDAR 2017)'
+Paper:
+  Title: ICDAR2017 Robust Reading Challenge on Multi-Lingual Scene Text Detection and Script Identification - RRC-MLT
+  URL: https://ieeexplore.ieee.org/document/8270168
+  Venue: ICDAR
+  Year: '2017'
+  BibTeX: '@INPROCEEDINGS{8270168,
+  author={Nayef, Nibal and Yin, Fei and Bizid, Imen and Choi, Hyunsoo and Feng, Yuan and Karatzas, Dimosthenis and Luo, Zhenbo and Pal, Umapada and Rigaud, Christophe and Chazalon, Joseph and Khlif, Wafa and Luqman, Muhammad Muzzamil and Burie, Jean-Christophe and Liu, Cheng-lin and Ogier, Jean-Marc},
+  booktitle={2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)},
+  title={ICDAR2017 Robust Reading Challenge on Multi-Lingual Scene Text Detection and Script Identification - RRC-MLT},
+  year={2017},
+  volume={01},
+  number={},
+  pages={1454-1459},
+  doi={10.1109/ICDAR.2017.237}}'
+Data:
+  Website: https://rrc.cvc.uab.es/?ch=8
+  Language:
+    - Arabic
+    - English
+    - Chinese
+    - Japanese
+    - Korean
+    - Italian
+    - German
+    - Indian
+    - French
+  Scene:
+    - Natural Scene
+  Granularity:
+    - Word
+  Tasks:
+    - textdet
+    - textrecog
+    - textspotting
+  License:
+    Type: CC BY 4.0
+    Link: https://creativecommons.org/licenses/by/4.0/
+  Format: .txt

dataset_zoo/mlt/sample_anno.md

@@ -0,0 +1,20 @@
+**Text Detection, Text Spotting**
+
+```text
+# x1,y1,x2,y2,x3,y3,x4,y4,script,text
+# Valid scripts are: "Arabic", "Latin", "Chinese", "Japanese", "Korean", "Bangla", "Symbols", "Mixed", "None"
+
+131,34,181,34,179,47,131,49,Latin,Castle
+150,59,194,58,196,72,150,73,Arabic,متحف
+90,83,143,83,143,96,91,96,Latin,Heritage
+146,81,200,80,201,93,147,94,Latin,Museum
+```
+
+**Text Recognition**
+
+```text
+# img_name,script,text
+
+word_4.png,Arabic,المكرمة
+word_5.png,Latin,MAKKA
+```

dataset_zoo/mlt/textdet.py

@@ -0,0 +1,114 @@
+data_root = 'data/mlt'
+cache_path = 'data/cache'
+# yapf: disable
+train_preparer = dict(
+    obtainer=dict(
+        type='NaiveDataObtainer',
+        cache_path=cache_path,
+        files=[
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_1.zip',  # noqa: E501
+                save_name='mlt_1.zip',
+                md5='7b26e10d949c00fb4411f40b4f1fce6e',
+                content=['image'],
+                mapping=[['mlt_1/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_2.zip',  # noqa: E501
+                save_name='mlt_2.zip',
+                md5='e992fb5a7621dd6329081a73e52a28e1',
+                content=['image'],
+                mapping=[['mlt_2/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_3.zip',  # noqa: E501
+                save_name='mlt_3.zip',
+                md5='044ea5fb1dcec8bbb874391c517b55ff',
+                content=['image'],
+                mapping=[['mlt_3/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_4.zip',  # noqa: E501
+                save_name='mlt_4.zip',
+                md5='344a657c1cc7cbb150547f1c76b5cc8e',
+                content=['image'],
+                mapping=[['mlt_4/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_5.zip',  # noqa: E501
+                save_name='mlt_5.zip',
+                md5='5c7ac0158e7189c0a634eaf7bdededc5',
+                content=['image'],
+                mapping=[['mlt_5/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_6.zip',  # noqa: E501
+                save_name='mlt_6.zip',
+                md5='3b479255a96d255680f51005b5232bac',
+                content=['image'],
+                mapping=[['mlt_6/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_7.zip',  # noqa
+                save_name='mlt_7.zip',
+                md5='faa033fb9d2922d747bad9b0692c992e',
+                content=['image'],
+                mapping=[['mlt_7/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/ch8_training_images_8.zip',  # noqa
+                save_name='mlt_8.zip',
+                md5='db8afa59ae520757151f6ce5acd489ef',
+                content=['image'],
+                mapping=[['mlt_8/*', 'textdet_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_training_localization_transcription_gt_v2.zip',
+                save_name='mlt_train_gt.zip',
+                md5='2c9c3de30b5615f6846738bbd336c988',
+                content=['annotation'],
+                mapping=[['mlt_train_gt/', 'annotations/train']]),
+        ]),
+    gatherer=dict(
+        type='PairGatherer',
+        img_suffixes=['.jpg', '.JPG'],
+        rule=[r'img_(\d+)\.([jJ][pP][gG])', r'gt_img_\1.txt']),
+    parser=dict(
+        type='ICDARTxtTextDetAnnParser',
+        encoding='utf-8-sig',
+        format='x1,y1,x2,y2,x3,y3,x4,y4,lang,trans'),
+    packer=dict(type='TextDetPacker'),
+    dumper=dict(type='JsonDumper'),
+)  # noqa
+
+val_preparer = dict(
+    obtainer=dict(
+        type='NaiveDataObtainer',
+        cache_path=cache_path,
+        files=[
+            dict(
+                url='https://rrc.cvc.uab.es/downloads/ch8_validation_images.zip',  # noqa
+                save_name='mlt_val_img.zip',
+                md5='3cfc7b440ab81b89a981d707786dbe83',
+                content=['image'],
+                mapping=[['mlt_val_img', 'textdet_imgs/val']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_validation_localization_transcription_gt_v2.zip',
+                save_name='mlt_val_gt.zip',
+                md5='ecae7d433e6f103bb31e00d37254009c',
+                content=['annotation'],
+                mapping=[['mlt_val_gt/', 'annotations/val']]),
+        ]),
+    gatherer=dict(
+        type='PairGatherer',
+        img_suffixes=['.jpg', '.JPG'],
+        rule=[r'img_(\d+)\.([jJ][pP][gG])', r'gt_img_\1.txt']),
+    parser=dict(
+        type='ICDARTxtTextDetAnnParser',
+        encoding='utf-8-sig',
+        format='x1,y1,x2,y2,x3,y3,x4,y4,lang,trans'),
+    packer=dict(type='TextDetPacker'),
+    dumper=dict(type='JsonDumper'),
+)
+
+config_generator = dict(
+    type='TextDetConfigGenerator',
+    val_anns=[dict(ann_file='textdet_val.json', dataset_postfix='')],
+    test_anns=None)
+
+delete = [f'mlt{i}' for i in range(1, 9)
+          ] + ['annotations', 'mlt_val_gt', 'mlt_train_gt']

dataset_zoo/mlt/textrecog.py

@@ -0,0 +1,82 @@
+data_root = 'data/mlt'
+cache_path = 'data/cache'
+train_preparer = dict(
+    obtainer=dict(
+        type='NaiveDataObtainer',
+        cache_path=cache_path,
+        files=[
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_training_word_images_gt_part_1.zip',
+                save_name='mlt_rec_1.zip',
+                md5='714d899cf5c8cf23b73bc14cfb628a3a',
+                content=['image'],
+                mapping=[['mlt_rec_1/*', 'textrecog_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_training_word_images_gt_part_2.zip',
+                save_name='mlt_rec_2.zip',
+                md5='d0e5bc4736626853203d24c70bbf56d1',
+                content=['image'],
+                mapping=[['mlt_rec_2/*', 'textrecog_imgs/train']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_training_word_images_gt_part_3.zip',
+                save_name='mlt_rec_3.zip',
+                md5='ebc7f2c9e73c3d174437d43b03177c5c',
+                content=['image'],
+                mapping=[['mlt_rec_3/*', 'textrecog_imgs/train']]),
+            dict(
+                url='https://rrc.cvc.uab.es/downloads/'
+                'ch8_validation_word_images_gt.zip',
+                save_name='mlt_rec_train_gt.zip',
+                md5='e5e681b440a616f0ac8deaa669b3682d',
+                content=['annotation'],
+                mapping=[['mlt_rec_train_gt/', 'annotations/train']]),
+        ]),
+    gatherer=dict(type='MonoGatherer', ann_name='train/gt.txt'),
+    parser=dict(
+        type='ICDARTxtTextRecogAnnParser',
+        encoding='utf-8-sig',
+        format='img,lang,text'),
+    packer=dict(type='TextRecogPacker'),
+    dumper=dict(type='JsonDumper'),
+)
+
+val_preparer = dict(
+    obtainer=dict(
+        type='NaiveDataObtainer',
+        cache_path=cache_path,
+        files=[
+            dict(
+                url='https://rrc.cvc.uab.es/downloads/'
+                'ch8_validation_word_images_gt.zip',
+                save_name='mlt_rec_val.zip',
+                md5='954acd0325c442288fa4aff1009b6d79',
+                content=['image'],
+                mapping=[['mlt_rec_val/*', 'textrecog_imgs/val']]),
+            dict(
+                url='https://datasets.cvc.uab.es/rrc/'
+                'ch8_validation_word_gt_v2.zip',
+                save_name='mlt_rec_val_gt.zip',
+                md5='951c9cee78a0064b133ab59369a9b232',
+                content=['annotation'],
+                mapping=[['mlt_rec_val_gt/', 'annotations/val']]),
+        ]),
+    gatherer=dict(type='MonoGatherer', ann_name='train/gt.txt'),
+    parser=dict(
+        type='ICDARTxtTextRecogAnnParser',
+        encoding='utf-8-sig',
+        format='img,lang,text'),
+    packer=dict(type='TextRecogPacker'),
+    dumper=dict(type='JsonDumper'),
+)
+
+config_generator = dict(
+    type='TextRecogConfigGenerator',
+    val_anns=[dict(ann_file='textrecog_val.json', dataset_postfix='')],
+    test_anns=None)
+
+delete = [f'mlt_rec_{i}' for i in range(1, 4)] + [
+    'annotations', 'mlt_rec_val_gt', 'mlt_rec_train_gt', 'mlt_rec_val'
+]

dataset_zoo/mlt/textspotting.py

@@ -0,0 +1,9 @@
+_base_ = ['textdet.py']
+
+_base_.train_preparer.gatherer.img_dir = 'textdet_imgs/train'
+_base_.train_preparer.packer.type = 'TextSpottingPacker'
+
+_base_.val_preparer.gatherer.img_dir = 'textdet_imgs/val'
+_base_.val_preparer.packer.type = 'TextSpottingPacker'
+
+config_generator = dict(type='TextSpottingConfigGenerator')

mmocr/datasets/preparers/obtainers/naive_data_obtainer.py

@@ -177,22 +177,32 @@ def move(self, mapping: List[Tuple[str, str]]) -> None:
 
         Args:
             mapping (List[Tuple[str, str]]): A list of tuples, each
-            tuple contains the source file name and the destination file name.
+                tuple contains the source file name and the destination file
+                name.
         """
         for src, dst in mapping:
             src = osp.join(self.data_root, src)
             dst = osp.join(self.data_root, dst)
 
             if '*' in src:
+                # dst must be a directory
                 mkdir_or_exist(dst)
                 for f in glob.glob(src):
-                    if not osp.exists(
-                            osp.join(dst, osp.relpath(f, self.data_root))):
+                    tgt = osp.join(dst, osp.basename(osp.normpath(f)))
+                    if not osp.exists(tgt):
                         shutil.move(f, dst)
-
-            elif osp.exists(src) and not osp.exists(dst):
-                mkdir_or_exist(osp.dirname(dst))
-                shutil.move(src, dst)
+                    else:
+                        print(f'Skipping moving {f} to {dst} since'
+                              f' {f} does not exist or {tgt} already exists')
+            # If no wildcard in src, dst must match the src type
+            # That is, we can only move a file to a file, or a dir to a dir
+            else:
+                if osp.exists(src) and not osp.exists(dst):
+                    mkdir_or_exist(osp.dirname(dst))
+                    shutil.move(src, dst)
+                else:
+                    print(f'Skipping moving {src} to {dst} since'
+                          f' {src} does not exist or {dst} already exists')
 
     def clean(self) -> None:
         """Remove empty dirs."""

mmocr/datasets/preparers/parsers/icdar_txt_parser.py

@@ -20,7 +20,8 @@ class ICDARTxtTextDetAnnParser(BaseParser):
             to ','.
         ignore (str): The text to be ignored. Defaults to '###'.
         format (str): The format of the annotation. Defaults to
-            'x1,y1,x2,y2,x3,y3,x4,trans'.
+            'x1,y1,x2,y2,x3,y3,x4,trans'. An additional keyword "lang" can be
+            recognized here to specify the language of the transcription.
         encoding (str): The encoding of the annotation file. Defaults to
             'utf-8-sig'.
         nproc (int): The number of processes to parse the annotation. Defaults
@@ -52,6 +53,8 @@ def parse_file(self, img_path: str, ann_path: str) -> Tuple:
         instances = list()
         for anno in self.loader(ann_path, self.sep, self.format,
                                 self.encoding):
+            if 'lang' in anno:
+                del anno['lang']
             anno = list(anno.values())
             if self.remove_strs is not None:
                 for strs in self.remove_strs:
@@ -82,6 +85,8 @@ class ICDARTxtTextRecogAnnParser(BaseParser):
             to ','.
         ignore (str): The text to be ignored. Defaults to '#'.
         format (str): The format of the annotation. Defaults to 'img, text'.
+            An additional keyword "lang" can be recognized here to specify the
+            language of the transcription.
         encoding (str): The encoding of the annotation file. Defaults to
             'utf-8-sig'.
         nproc (int): The number of processes to parse the annotation. Defaults