Data pipeline refactoring

.flake8

@@ -21,7 +21,7 @@ per-file-ignores =
     tools/infer/text/predict_rec.py:E402
     tools/dataset_converters/convert.py:F401,F403
     deploy/models_utils/auto_scaling/converter.py:E402
-    mindocr/data/transforms/transforms_factory.py:F401,F403
+    mindocr/data/transforms/transforms_factory.py:F401,F403,F405
     tests/*:E402
     tests/ut/dataset_convert/test_dataset_converter.py:E402
     tests/ut/test_datasets.py:F401,E402

configs/det/dbnet/db_mobilenetv3_icdar15.yaml

@@ -32,7 +32,7 @@ model:
 
 postprocess:
   name: DBPostprocess
-  box_type: quad   # whether to output a polygon or a box
+  box_type: quad          # whether to output a polygon or a box
   binary_thresh: 0.3      # binarization threshold
   box_thresh: 0.6         # box score threshold
   max_candidates: 1000
@@ -79,13 +79,8 @@ train:
     label_file: ic15/det/train/det_gt.txt
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
-      - RandomColorAdjust:
-          brightness: 0.1255  # 32.0 / 255
-          saturation: 0.5
       - RandomHorizontalFlip:
           p: 0.5
       - RandomRotate:
@@ -108,15 +103,15 @@ train:
           shrink_ratio: 0.4
           thresh_min: 0.3
           thresh_max: 0.7
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - RandomColorAdjust:
+          brightness: 0.1255  # 32.0 / 255
+          saturation: 0.5
+      - Normalize:
+          mean: &mean [ 123.675, 116.28, 103.53 ]
+          std: &std [ 58.395, 57.12, 57.375 ]
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visualize
-    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask']
-#    output_columns: ['image'] # for debug op performance
+    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ]
     net_input_column_index: [0] # input indices for network forward func in output_columns
     label_column_index: [1, 2, 3, 4] # input indices marked as label
 
@@ -136,22 +131,16 @@ eval:
     label_file: ic15/det/test/det_gt.txt
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
-      - GridResize:
-          factor: 32
-      # GridResize already sets the evaluation size to [ 736, 1280 ].
-      # Uncomment ScalePadImage block for other resolutions.
-#      - ScalePadImage:
-#          target_size: [ 736, 1280 ] # h, w
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - DetResize:
+          target_size: [ 736, 1280 ]
+          keep_ratio: False
+          force_divisible: True #  GridResize 32
+      - Normalize:
+          mean: *mean
+          std: *std
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the labels for evaluation
     output_columns: [ 'image', 'polys', 'ignore_tags', 'shape_list' ]
     net_input_column_index: [0] # input indices for network forward func in output_columns

configs/det/dbnet/db_r18_icdar15.yaml

@@ -26,7 +26,7 @@ model:
 
 postprocess:
   name: DBPostprocess
-  box_type: quad   # whether to output a polygon or a box
+  box_type: quad          # whether to output a polygon or a box
   binary_thresh: 0.3      # binarization threshold
   box_thresh: 0.6         # box score threshold
   max_candidates: 1000
@@ -68,18 +68,14 @@ train:
   dataset_sink_mode: True
   dataset:
     type: DetDataset
+    mindrecord: True
     dataset_root: /data/ocr_datasets
-    data_dir: ic15/det/train/ch4_training_images
-    label_file: ic15/det/train/det_gt.txt
+    data_dir: ic15/det/MR/train
+    label_file: ''
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
-      - RandomColorAdjust:
-          brightness: 0.1255  # 32.0 / 255
-          saturation: 0.5
       - RandomHorizontalFlip:
           p: 0.5
       - RandomRotate:
@@ -102,15 +98,15 @@ train:
           shrink_ratio: 0.4
           thresh_min: 0.3
           thresh_max: 0.7
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - RandomColorAdjust:
+          brightness: 0.1255  # 32.0 / 255
+          saturation: 0.5
+      - Normalize:
+          mean: &mean [ 123.675, 116.28, 103.53 ]
+          std: &std [ 58.395, 57.12, 57.375 ]
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visualize
-    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ] #'img_path']
-#    output_columns: ['image'] # for debug op performance
+    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ]
     net_input_column_index: [0] # input indices for network forward func in output_columns
     label_column_index: [1, 2, 3, 4] # input indices marked as label
 
@@ -125,24 +121,21 @@ eval:
   dataset_sink_mode: False
   dataset:
     type: DetDataset
+    mindrecord: True
     dataset_root: /data/ocr_datasets
-    data_dir: ic15/det/test/ch4_test_images
-    label_file: ic15/det/test/det_gt.txt
+    data_dir: ic15/det/MR/test
+    label_file: ''
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
       - DetResize:
           target_size: [ 736, 1280 ]
           keep_ratio: False
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - Normalize:
+          mean: *mean
+          std: *std
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the labels for evaluation
     output_columns: [ 'image', 'polys', 'ignore_tags', 'shape_list']
     net_input_column_index: [0] # input indices for network forward func in output_columns

configs/det/dbnet/db_r50_icdar15.yaml

@@ -25,7 +25,7 @@ model:
 
 postprocess:
   name: DBPostprocess
-  box_type: quad   # whether to output a polygon or a box
+  box_type: quad          # whether to output a polygon or a box
   binary_thresh: 0.3      # binarization threshold
   box_thresh: 0.6         # box score threshold
   max_candidates: 1000
@@ -67,18 +67,14 @@ train:
   dataset_sink_mode: True
   dataset:
     type: DetDataset
+    mindrecord: True
     dataset_root: /data/ocr_datasets
-    data_dir: ic15/det/train/ch4_training_images
-    label_file: ic15/det/train/det_gt.txt
+    data_dir: ic15/det/MR/train
+    label_file: ''
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
-      - RandomColorAdjust:
-          brightness: 0.1255  # 32.0 / 255
-          saturation: 0.5
       - RandomHorizontalFlip:
           p: 0.5
       - RandomRotate:
@@ -101,15 +97,15 @@ train:
           shrink_ratio: 0.4
           thresh_min: 0.3
           thresh_max: 0.7
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - RandomColorAdjust:
+          brightness: 0.1255  # 32.0 / 255
+          saturation: 0.5
+      - Normalize:
+          mean: &mean [ 123.675, 116.28, 103.53 ]
+          std: &std [ 58.395, 57.12, 57.375 ]
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visualize
-    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ] #'img_path']
-#    output_columns: ['image'] # for debug op performance
+    output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ]
     net_input_column_index: [0] # input indices for network forward func in output_columns
     label_column_index: [1, 2, 3, 4] # input indices marked as label
 
@@ -124,31 +120,22 @@ eval:
   dataset_sink_mode: False
   dataset:
     type: DetDataset
+    mindrecord: True
     dataset_root: /data/ocr_datasets
-    data_dir: ic15/det/test/ch4_test_images
-    label_file: ic15/det/test/det_gt.txt
+    data_dir: ic15/det/MR/test
+    label_file: ''
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
+      - Decode:
       - DetLabelEncode:
-      #- GridResize:
-      #    factor: 32
-      # GridResize already sets the evaluation size to [ 736, 1280 ].
-      # Uncomment ScalePadImage block for other resolutions.
-#      - ScalePadImage:
-#          target_size: [ 736, 1280 ] # h, w
       - DetResize:
           target_size: [ 736, 1280 ]
           keep_ratio: False
-          force_divisable: True #  GridResize 32
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+          force_divisible: True #  GridResize 32
+      - Normalize:
+          mean: *mean
+          std: *std
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the labels for evaluation
     output_columns: [ 'image', 'polys', 'ignore_tags', 'shape_list' ]
     net_input_column_index: [0] # input indices for network forward func in output_columns

configs/det/dbnet/db_r50_synthtext.yaml

@@ -60,12 +60,7 @@ train:
     label_file: SynthText/gt_processed.mat
     sample_ratio: 1.0
     transform_pipeline:
-      - DecodeImage:
-          img_mode: RGB
-          to_float32: False
-      - RandomColorAdjust:
-          brightness: 0.1255  # 32.0 / 255
-          saturation: 0.5
+      - Decode:
       - RandomHorizontalFlip:
           p: 0.5
       - RandomRotate:
@@ -88,12 +83,13 @@ train:
           shrink_ratio: 0.4
           thresh_min: 0.3
           thresh_max: 0.7
-      - NormalizeImage:
-          bgr_to_rgb: False
-          is_hwc: True
-          mean: imagenet
-          std: imagenet
-      - ToCHWImage:
+      - RandomColorAdjust:
+          brightness: 0.1255  # 32.0 / 255
+          saturation: 0.5
+      - Normalize:
+          mean: [ 123.675, 116.28, 103.53 ]
+          std: [ 58.395, 57.12, 57.375 ]
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visualize
     output_columns: [ 'image', 'binary_map', 'mask', 'thresh_map', 'thresh_mask' ] #'img_path']
 #    output_columns: ['image'] # for debug op performance

configs/det/east/east_r50_icdar15.yaml

@@ -111,7 +111,7 @@ eval:
       - DetLabelEncode:
       - DetResize:
           target_size: [720, 1280]
-          force_divisable: False
+          force_divisible: False
       - NormalizeImage:
           bgr_to_rgb: False
           is_hwc: True

configs/rec/crnn/crnn_icdar15.yaml

@@ -106,7 +106,7 @@ train:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_seq'] #, 'length'] #'img_path']
     net_input_column_index: [0] # input indices for network forward func in output_columns
@@ -147,7 +147,7 @@ eval:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_padded', 'text_length']  # TODO return text string padding w/ fixed length, and a scaler to indicate the length
     net_input_column_index: [0] # input indices for network forward func in output_columns

configs/rec/crnn/crnn_resnet34.yaml

@@ -97,7 +97,7 @@ train:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_seq'] #, 'length'] #'img_path']
     net_input_column_index: [0] # input indices for network forward func in output_columns
@@ -140,7 +140,7 @@ eval:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_padded', 'text_length']  # TODO return text string padding w/ fixed length, and a scaler to indicate the length
     net_input_column_index: [0] # input indices for network forward func in output_columns

configs/rec/crnn/crnn_vgg7.yaml

@@ -98,7 +98,7 @@ train:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_seq'] #, 'length'] #'img_path']
     net_input_column_index: [0] # input indices for network forward func in output_columns
@@ -141,7 +141,7 @@ eval:
           is_hwc: True
           mean : [127.0, 127.0, 127.0]
           std : [127.0, 127.0, 127.0]
-      - ToCHWImage:
+      - HWC2CHW:
     #  the order of the dataloader list, matching the network input and the input labels for the loss function, and optional data for debug/visaulize
     output_columns: ['image', 'text_padded', 'text_length']  # TODO return text string padding w/ fixed length, and a scaler to indicate the length
     net_input_column_index: [0] # input indices for network forward func in output_columns