feat: Add model configs and demo ipynb of tetradi dataset model

tetradi/ocr_config.json

@@ -0,0 +1,76 @@
+{
+    "alphabet": " !\"'()*+,-./0123456789:;<=>?ABCDEFGHIJKLMNOPRSTVWY[\\]_abcdefghiklmnoprstuvwxyz|}ЁАБВГДЕЖЗИКЛМНОПРСТУФХЦЧШЩЫЬЭЮЯабвгдежзийклмнопрстуфхцчшщъыьэюяё’№",
+    "save_dir": "/home/jovyan/skalinin/OCR-model/data/experiments/1024_128_add_new_dataset_album_augs_2",
+    "num_epochs": 100,
+    "pretrain_path": "",
+    "image": {
+        "width": 1024,
+        "height": 128
+    },
+    "train": {
+        "datasets": [
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/train.csv",
+                "prob": 0.05
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/train.csv",
+                "prob": 0.1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/train.csv",
+                "prob": 0.45
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/train.csv",
+                "prob": 0.4
+            }
+        ],
+        "epoch_size": 100000,
+        "batch_size": 64
+    },
+    "val": {
+        "datasets": [
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/val.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/val.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/val.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/val.csv",
+                "prob": 1
+            }
+        ],
+        "epoch_size": null,
+        "batch_size": 64
+    },
+    "test": {
+        "datasets": [
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_en/test.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_11032022_ru/test.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_15102021/test.csv",
+                "prob": 1
+            },
+            {
+                "csv_path": "/home/jovyan/skalinin/OCR-model/data/segmentation_26012022/test.csv",
+                "prob": 1
+            }
+        ],
+        "epoch_size": null,
+        "batch_size": 64
+    }
+}

tetradi/pipeline_config.json

@@ -0,0 +1,46 @@
+{   "main_process": {
+        "ImageToBGR": {
+            "input_format": "BGR"
+        },
+        "SegmPrediction": {
+            "model_path": "model-319-0.1578.ckpt",
+            "config_path": "segm_config-fitting_text_line_11032022.json",
+            "device": "cuda"
+        },
+        "RestoreImageAngle": {
+            "restoring_class_names": ["text_line"]
+        },
+        "ClassContourPosptrocess": {},
+        "OCRPrediction": {
+            "model_path": "model-97-0.8179.ckpt",
+            "config_path": "ocr_config.json",
+            "lm_path": "corpus_o6.arpa",
+            "classes_to_ocr": ["shrinked_pupil_text", "shrinked_comments"],
+            "device": "cuda"
+        },
+        "LineFinder": {
+            "line_classes": ["text_line"],
+            "text_classes": ["shrinked_pupil_text", "shrinked_comments"]
+        },
+        "PrepareJSON": {}
+    },
+    "classes": {
+        "shrinked_pupil_text": {
+            "contour_posptrocess": {
+                "BboxFromContour": {},
+                "UpscaleBbox": {"upscale_bbox": [1.4, 2.3]},
+                "CropByBbox": {}
+            }
+        },
+        "shrinked_comments": {
+            "contour_posptrocess": {
+                "BboxFromContour": {},
+                "UpscaleBbox": {"upscale_bbox": [1.4, 2.3]},
+                "CropByBbox": {}
+            }
+        },
+        "text_line": {
+            "contour_posptrocess": {}
+        }
+    }
+}

tetradi/segm_config-fitting_text_line_11032022.json

@@ -0,0 +1,72 @@
+{
+    "save_dir": "/home/jovyan/SEGM-model/data/exp/fitting_text_line_11032022",
+    "num_epochs": 1000,
+    "pretrain_path": "/home/jovyan/OCR-pipeline/data/weights/resnet50_15102021_26012022_768_1024_2/model-1249-0.1312.ckpt",
+    "image": {
+        "width": 1024,
+        "height": 768
+    },
+    "classes": {
+        "shrinked_pupil_text": {
+            "annotation_classes": ["pupil_text"],
+            "polygon2mask": {
+                "ShrinkMaskMaker": {"shrink_ratio": 0.5}
+            },
+            "postprocess": {
+                "threshold": 0.8,
+                "min_area": 10
+            }
+        },
+        "shrinked_comments": {
+            "annotation_classes": ["pupil_comment", "teacher_comment"],
+            "polygon2mask": {
+                "ShrinkMaskMaker": {"shrink_ratio": 0.5}
+            },
+            "postprocess": {
+                "threshold": 0.8,
+                "min_area": 10
+            }
+        },
+        "text_line": {
+            "annotation_classes": ["text_line"],
+            "polygon2mask": {
+                "PolylineToMask": {"thickness": 2, "scale": 0.95}
+            },
+            "postprocess": {
+                "threshold": 0.8,
+                "min_area": 10
+            }
+        }
+    },
+    "train": {
+        "datasets": [
+            {
+                "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_train.json",
+                "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
+                "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_train.csv"
+            }
+        ],
+        "batch_size": 10
+    },
+    "val": {
+        "datasets": [
+            {
+                "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_val.json",
+                "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
+                "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_val.csv"
+            }
+        ],
+        "batch_size": 10
+    },
+    "test": {
+        "datasets": [
+            {
+                "json_path": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/annotations_test.json",
+                "image_root": "/home/jovyan/data/htr_dataset_SEGM_11032022/en_ru_tetradi/ru/images_rotated/",
+                "processed_data_path": "/home/jovyan/SEGM-model/data/htr_dataset_SEGM_11032022_ru/annotations_val.csv"
+            }
+        ],
+        "batch_size": 10
+    }
+}
+