Stanislav Kalinin commited on
Commit
ce41b2a
1 Parent(s): 2c7c0ba

feat: Update school notebooks models

Browse files
ocr/ocr_config.json CHANGED
@@ -1,76 +1,52 @@
1
  {
2
  "alphabet": " !\"'()*+,-./0123456789:;<=>?ABCDEFGHIJKLMNOPRSTVWY[\\]_abcdefghiklmnoprstuvwxyz|}ЁАБВГДЕЖЗИКЛМНОПРСТУФХЦЧШЩЫЬЭЮЯабвгдежзийклмнопрстуфхцчшщъыьэюяё’№",
3
- "save_dir": "/home/jovyan/skalinin/OCR-model/data/experiments/1024_128_add_new_dataset_album_augs_2",
4
- "num_epochs": 100,
5
- "pretrain_path": "",
6
  "image": {
7
- "width": 1024,
8
- "height": 128
9
  },
10
  "train": {
11
  "datasets": [
12
  {
13
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/train.csv",
14
- "prob": 0.05
15
  },
16
  {
17
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/train.csv",
18
- "prob": 0.1
19
- },
20
- {
21
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/train.csv",
22
- "prob": 0.45
23
- },
24
- {
25
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/train.csv",
26
- "prob": 0.4
27
  }
28
  ],
29
  "epoch_size": 100000,
30
- "batch_size": 64
31
  },
32
  "val": {
33
  "datasets": [
34
  {
35
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/val.csv",
36
- "prob": 1
37
- },
38
- {
39
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/val.csv",
40
- "prob": 1
41
  },
42
  {
43
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/val.csv",
44
- "prob": 1
45
- },
46
- {
47
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/val.csv",
48
- "prob": 1
49
  }
50
  ],
51
  "epoch_size": null,
52
- "batch_size": 64
53
  },
54
  "test": {
55
  "datasets": [
56
  {
57
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/test.csv",
58
- "prob": 1
59
- },
60
- {
61
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/test.csv",
62
- "prob": 1
63
- },
64
- {
65
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/test.csv",
66
- "prob": 1
67
  },
68
  {
69
- "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/test.csv",
70
- "prob": 1
71
  }
72
  ],
73
  "epoch_size": null,
74
- "batch_size": 64
75
  }
76
  }
 
1
  {
2
  "alphabet": " !\"'()*+,-./0123456789:;<=>?ABCDEFGHIJKLMNOPRSTVWY[\\]_abcdefghiklmnoprstuvwxyz|}ЁАБВГДЕЖЗИКЛМНОПРСТУФХЦЧШЩЫЬЭЮЯабвгдежзийклмнопрстуфхцчшщъыьэюяё’№",
3
+ "save_dir": "data/experiments/tetradi_64_512_exp2/",
4
+ "num_epochs": 300,
5
+ "pretrain_path": "data/pretrain/model-187-0.6814.ckpt",
6
  "image": {
7
+ "width": 512,
8
+ "height": 64
9
  },
10
  "train": {
11
  "datasets": [
12
  {
13
+ "csv_path": "data/school_notebooks_EN/train/train.csv",
14
+ "prob": 14595
15
  },
16
  {
17
+ "csv_path": "data/school_notebooks_RU/train/train.csv",
18
+ "prob": 268350
 
 
 
 
 
 
 
 
19
  }
20
  ],
21
  "epoch_size": 100000,
22
+ "batch_size": 512
23
  },
24
  "val": {
25
  "datasets": [
26
  {
27
+ "csv_path": "data/school_notebooks_EN/val/val.csv",
28
+ "prob": 2129
 
 
 
 
29
  },
30
  {
31
+ "csv_path": "data/school_notebooks_RU/val/val.csv",
32
+ "prob": 27853
 
 
 
 
33
  }
34
  ],
35
  "epoch_size": null,
36
+ "batch_size": 512
37
  },
38
  "test": {
39
  "datasets": [
40
  {
41
+ "csv_path": "data/school_notebooks_EN/test/test.csv",
42
+ "prob": 1979
 
 
 
 
 
 
 
 
43
  },
44
  {
45
+ "csv_path": "data/school_notebooks_RU/test/test.csv",
46
+ "prob": 28109
47
  }
48
  ],
49
  "epoch_size": null,
50
+ "batch_size": 512
51
  }
52
  }
ocr/ocr_model.ckpt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9bd612eeb20a39d3cf83d10e89ed8b06ee844143e60607d28af218abd8ebc297
3
- size 50294787
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74192c3c23532f8d655c806d09aeeccf88b7e9fb769a21579f724702b99771ec
3
+ size 50289625
ocr/ocr_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc5ada98e82336c3e229b78aff6758cda6c92e6ca2cae5d9118b3e447994fd6e
3
+ size 50175647
pipeline_config.json CHANGED
@@ -5,7 +5,9 @@
5
  "SegmPrediction": {
6
  "model_path": "segm/segm_model.ckpt",
7
  "config_path": "segm/segm_config.json",
8
- "device": "cuda"
 
 
9
  },
10
  "RestoreImageAngle": {
11
  "restoring_class_names": ["text_line"]
@@ -15,24 +17,20 @@
15
  "model_path": "ocr/ocr_model.ckpt",
16
  "config_path": "ocr/ocr_config.json",
17
  "lm_path": "ocr/kenlm_corpus.arpa",
18
- "classes_to_ocr": ["shrinked_pupil_text", "shrinked_comments"],
19
- "device": "cuda"
 
 
 
20
  },
21
  "LineFinder": {
22
  "line_classes": ["text_line"],
23
- "text_classes": ["shrinked_pupil_text", "shrinked_comments"]
24
  },
25
  "PrepareJSON": {}
26
  },
27
  "classes": {
28
- "shrinked_pupil_text": {
29
- "contour_posptrocess": {
30
- "BboxFromContour": {},
31
- "UpscaleBbox": {"upscale_bbox": [1.4, 2.3]},
32
- "CropByBbox": {}
33
- }
34
- },
35
- "shrinked_comments": {
36
  "contour_posptrocess": {
37
  "BboxFromContour": {},
38
  "UpscaleBbox": {"upscale_bbox": [1.4, 2.3]},
 
5
  "SegmPrediction": {
6
  "model_path": "segm/segm_model.ckpt",
7
  "config_path": "segm/segm_config.json",
8
+ "num_threads": 8,
9
+ "device": "cuda",
10
+ "runtime": "Pytorch"
11
  },
12
  "RestoreImageAngle": {
13
  "restoring_class_names": ["text_line"]
 
17
  "model_path": "ocr/ocr_model.ckpt",
18
  "config_path": "ocr/ocr_config.json",
19
  "lm_path": "ocr/kenlm_corpus.arpa",
20
+ "num_threads": 8,
21
+ "classes_to_ocr": ["shrinked_text"],
22
+ "device": "cuda",
23
+ "batch_size": 128,
24
+ "runtime": "Pytorch"
25
  },
26
  "LineFinder": {
27
  "line_classes": ["text_line"],
28
+ "text_classes": ["shrinked_text"]
29
  },
30
  "PrepareJSON": {}
31
  },
32
  "classes": {
33
+ "shrinked_text": {
 
 
 
 
 
 
 
34
  "contour_posptrocess": {
35
  "BboxFromContour": {},
36
  "UpscaleBbox": {"upscale_bbox": [1.4, 2.3]},
segm/segm_config.json CHANGED
@@ -1,14 +1,14 @@
1
  {
2
- "save_dir": "/home/jovyan/SEGM-model/data/exp/fitting_text_line_11032022",
3
- "num_epochs": 1000,
4
- "pretrain_path": "/home/jovyan/OCR-pipeline/data/weights/resnet50_15102021_26012022_768_1024_2/model-1249-0.1312.ckpt",
5
  "image": {
6
- "width": 1024,
7
- "height": 768
8
  },
9
  "classes": {
10
- "shrinked_pupil_text": {
11
- "annotation_classes": ["pupil_text"],
12
  "polygon2mask": {
13
  "ShrinkMaskMaker": {"shrink_ratio": 0.5}
14
  },
@@ -17,10 +17,10 @@
17
  "min_area": 10
18
  }
19
  },
20
- "shrinked_comments": {
21
- "annotation_classes": ["pupil_comment", "teacher_comment"],
22
  "polygon2mask": {
23
- "ShrinkMaskMaker": {"shrink_ratio": 0.5}
24
  },
25
  "postprocess": {
26
  "threshold": 0.8,
@@ -30,7 +30,7 @@
30
  "text_line": {
31
  "annotation_classes": ["text_line"],
32
  "polygon2mask": {
33
- "PolylineToMask": {"thickness": 2, "scale": 0.95}
34
  },
35
  "postprocess": {
36
  "threshold": 0.8,
@@ -41,32 +41,55 @@
41
  "train": {
42
  "datasets": [
43
  {
44
- "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_train.json",
45
- "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
46
- "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_train.csv"
 
 
 
 
 
 
 
47
  }
48
  ],
49
- "batch_size": 10
 
50
  },
51
  "val": {
52
  "datasets": [
53
  {
54
- "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_val.json",
55
- "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
56
- "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_val.csv"
 
 
 
 
 
 
 
57
  }
58
  ],
59
- "batch_size": 10
 
60
  },
61
  "test": {
62
  "datasets": [
63
  {
64
- "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_test.json",
65
- "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
66
- "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_val.csv"
 
 
 
 
 
 
 
67
  }
68
  ],
69
- "batch_size": 10
 
70
  }
71
  }
72
-
 
1
  {
2
+ "save_dir": "data/experiments/tetradi_with_lines/",
3
+ "num_epochs": 1500,
4
+ "pretrain_path": "data/experiments/base_model_all_datasets/model-960-0.1979.ckpt",
5
  "image": {
6
+ "width": 896,
7
+ "height": 896
8
  },
9
  "classes": {
10
+ "shrinked_text": {
11
+ "annotation_classes": ["pupil_text", "teacher_comment"],
12
  "polygon2mask": {
13
  "ShrinkMaskMaker": {"shrink_ratio": 0.5}
14
  },
 
17
  "min_area": 10
18
  }
19
  },
20
+ "bordered_text": {
21
+ "annotation_classes": ["pupil_text", "teacher_comment"],
22
  "polygon2mask": {
23
+ "BorderMaskMaker": {"shrink_ratio": 0.5}
24
  },
25
  "postprocess": {
26
  "threshold": 0.8,
 
30
  "text_line": {
31
  "annotation_classes": ["text_line"],
32
  "polygon2mask": {
33
+ "PolylineToMask": {"thickness": 2}
34
  },
35
  "postprocess": {
36
  "threshold": 0.8,
 
41
  "train": {
42
  "datasets": [
43
  {
44
+ "json_path": "data/segm/school_notebooks_EN/annotations_train.json",
45
+ "image_root": "data/segm/school_notebooks_EN/images/",
46
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_EN/train/annotations_train.csv",
47
+ "prob": 70
48
+ },
49
+ {
50
+ "json_path": "data/segm/school_notebooks_RU/annotations_train.json",
51
+ "image_root": "data/segm/school_notebooks_RU/images/",
52
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_RU/train/annotations_train.csv",
53
+ "prob": 1557
54
  }
55
  ],
56
+ "epoch_size": 2000,
57
+ "batch_size": 20
58
  },
59
  "val": {
60
  "datasets": [
61
  {
62
+ "json_path": "data/segm/school_notebooks_EN/annotations_val.json",
63
+ "image_root": "data/segm/school_notebooks_EN/images/",
64
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_EN/val/annotations_val.csv",
65
+ "prob": 10
66
+ },
67
+ {
68
+ "json_path": "data/segm/school_notebooks_RU/annotations_val.json",
69
+ "image_root": "data/segm/school_notebooks_RU/images/",
70
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_RU/val/annotations_val.csv",
71
+ "prob": 150
72
  }
73
  ],
74
+ "epoch_size": null,
75
+ "batch_size": 20
76
  },
77
  "test": {
78
  "datasets": [
79
  {
80
+ "json_path": "data/segm/school_notebooks_EN/annotations_test.json",
81
+ "image_root": "data/segm/school_notebooks_EN/images/",
82
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_EN/test/annotations_test.csv",
83
+ "prob": 10
84
+ },
85
+ {
86
+ "json_path": "data/segm/school_notebooks_RU/annotations_test.json",
87
+ "image_root": "data/segm/school_notebooks_RU/images/",
88
+ "processed_data_path": "data/tetradi_with_lines/school_notebooks_RU/test/annotations_test.csv",
89
+ "prob": 150
90
  }
91
  ],
92
+ "epoch_size": null,
93
+ "batch_size": 20
94
  }
95
  }
 
segm/segm_model.ckpt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:47860d2f5d047a1b6f18bab61777544a7082227c32c6b58d28a4781af3d68784
3
- size 115723911
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12331d25804fd5b291707cbb288d384b81f7a6460e514b8695cbc3264491cde9
3
+ size 115714321
segm/segm_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:511b05e81ff544d236b7635b8a4367096cb679991c5d922d04c8fa7524f36127
3
+ size 115229451