Add ppocr series models

deploy/py_infer/src/configs/det/ppocr/ch_PP-OCRv2_det_cml.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: DistillationDBPostProcess
+  binary_thresh: 0.3
+  box_thresh: 0.6
+  max_candidates: 1000
+  expand_ratio: 1.5
+
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+      - ScalePadImage:
+      - NormalizeImage:
+          mean: [ 123.675, 116.28 , 103.53 ]
+          std: [ 58.395, 57.12 , 57.375 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/det/ppocr/ch_det_mv3_db_v2.0.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: DBPostprocess
+  binary_thresh: 0.3
+  box_thresh: 0.6
+  max_candidates: 1000
+  expand_ratio: 1.5
+
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+      - ScalePadImage:
+      - NormalizeImage:
+          mean: [ 123.675, 116.28 , 103.53 ]
+          std: [ 58.395, 57.12 , 57.375 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/arabic_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/ch_PP-OCRv2_rec_distillation.yaml

@@ -0,0 +1,15 @@
+postprocess:
+  name: DistillationCTCLabelDecode
+  use_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/chinese_cht_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/cyrillic_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/devanagari_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/en_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/japan_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/ka_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/korean_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/latin_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/rec_chinese_lite_train_v2.0.yaml

@@ -0,0 +1,16 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - CTCLabelEncode:
+      - RecResizeImg:
+          padding: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/rec_en_number_lite_train.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: True
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/ta_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/configs/rec/ppocr/te_PP-OCRv3_rec.yaml

@@ -0,0 +1,20 @@
+postprocess:
+  name: CTCLabelDecode
+  use_space_char: True
+  use_redundant_space_char: True
+  blank_at_last: False
+
+eval:
+  dataset:
+    transform_pipeline:
+      - DecodeImage:
+          img_mode: BGR
+          channel_first: False
+          to_float32: True
+      - RecResizeImg:
+          padding: False
+      - NormalizeImage:
+          mean: [ 127.5, 127.5, 127.5 ]
+          std: [ 127.5, 127.5, 127.5 ]
+          is_hwc: True
+      - ToCHWImage:

deploy/py_infer/src/data_process/postprocess/postprocess_mapping.py

@@ -3,5 +3,8 @@
 POSTPROCESS_MAPPING_OPS = {
     "DBPostprocess": db_postprocess.DBPostprocess,
     "RecCTCLabelDecode": rec_postprocess.RecCTCLabelDecode,
-    "ClsPostprocess": cls_postprocess.ClsPostprocess
+    "ClsPostprocess": cls_postprocess.ClsPostprocess,
+    "DistillationDBPostProcess": db_postprocess.DBPostprocess,
+    "DistillationCTCLabelDecode": rec_postprocess.RecCTCLabelDecode,
+    "CTCLabelDecode": rec_postprocess.RecCTCLabelDecode
 }

deploy/py_infer/src/data_process/postprocess/rec_postprocess.py

@@ -19,6 +19,7 @@ class RecCTCLabelDecode(object):
     def __init__(self,
                  character_dict_path=None,
                  use_space_char=False,
+                 use_redundant_space_char=False,
                  blank_at_last=True,
                  lower=False,
                  ):
@@ -48,6 +49,9 @@ def __init__(self,
                     "WARNING: The dict still contains space char in dict although use_space_char is set to be False, because the space char is coded in the dictionary file ",
                     character_dict_path)
 
+        if use_redundant_space_char:
+            char_list.append(' ')  # paddleOCR rec models need two space char
+
         self.num_valid_chars = len(char_list)  # the number of valid chars (including space char if used)
 
         # add blank token for padding

deploy/py_infer/src/data_process/preprocess/preprocess_mapping.py

@@ -1,6 +1,6 @@
 from ..preprocess import transforms
 
-PREPROCESS_SKIP_OPS = ["DetLabelEncode", "RecCTCLabelEncode"]
+PREPROCESS_SKIP_OPS = ["DetLabelEncode", "RecCTCLabelEncode", "CTCLabelEncode"]
 
 PREPROCESS_MAPPING_OPS = {
     "DecodeImage": transforms.DecodeImage,

deploy/py_infer/src/utils/adapted/paddleocr_models.py

@@ -4,9 +4,30 @@
 
 
 PADDLEOCR_MODELS = {
-    "ch_pp_server_det_v2.0": "det/ppocr/ch_det_res18_db_v2.0.yaml",            # ch_ppocr_server_v2.0_det
-    "ch_pp_det_v3": "det/ppocr/ch_PP-OCRv3_det_cml.yaml",                      # ch_PP-OCRv3_det
-    "ch_pp_server_rec_v2.0": "rec/ppocr/rec_chinese_common_train_v2.0.yaml",   # ch_ppocr_server_v2.0_rec
-    "ch_pp_rec_v3": "rec/ppocr/ch_PP-OCRv3_rec_distillation.yaml",             # ch_PP-OCRv3_rec
-    "ch_pp_mobile_cls_v2.0": "cls/ppocr/cls_mv3.yaml"                          # ch_ppocr_mobile_v2.0_cls
+    "ch_pp_server_det_v2.0": "det/ppocr/ch_det_res18_db_v2.0.yaml",                  # ch_ppocr_server_v2.0_det
+    "ch_pp_det_OCRv3": "det/ppocr/ch_PP-OCRv3_det_cml.yaml",                         # ch_PP-OCRv3_det
+    "ch_pp_server_rec_v2.0": "rec/ppocr/rec_chinese_common_train_v2.0.yaml",         # ch_ppocr_server_v2.0_rec
+    "ch_pp_rec_OCRv3": "rec/ppocr/ch_PP-OCRv3_rec_distillation.yaml",                # ch_PP-OCRv3_rec
+    "ch_pp_mobile_cls_v2.0": "cls/ppocr/cls_mv3.yaml",                               # ch_ppocr_mobile_v2.0_cls
+    "ch_pp_det_OCRv2": "det/ppocr/ch_PP-OCRv2_det_cml.yaml",                         # ch_PP-OCRv2_det
+    "ch_pp_mobile_det_v2.0_slim": "det/ppocr/ch_det_mv3_db_v2.0.yaml",               # ch_ppocr_mobile_slim_v2.0_det
+    "ch_pp_mobile_det_v2.0": "det/ppocr/ch_det_mv3_db_v2.0.yaml",                    # ch_ppocr_mobile_v2.0_det
+    "en_pp_det_OCRv3": "det/ppocr/ch_PP-OCRv3_det_cml.yaml",                         # en_PP-OCRv3_det
+    "ml_pp_det_OCRv3": "det/ppocr/ch_PP-OCRv3_det_cml.yaml",                         # ml_PP-OCRv3_det
+    "ch_pp_rec_OCRv2": "rec/ppocr/ch_PP-OCRv2_rec_distillation.yaml",                # ch_PP-OCRv2_rec
+    "ch_pp_mobile_rec_v2.0": "rec/ppocr/rec_chinese_lite_train_v2.0.yaml",           # ch_ppocr_mobile_v2.0_rec
+    "en_pp_rec_OCRv3": "rec/ppocr/en_PP-OCRv3_rec.yaml",                             # en_PP-OCRv3_rec
+    "en_pp_mobile_rec_number_v2.0_slim": "rec/ppocr/rec_en_number_lite_train.yaml",  # en_number_mobile_slim_v2.0_rec
+    "en_pp_mobile_rec_number_v2.0": "rec/ppocr/rec_en_number_lite_train.yaml",       # en_number_mobile_v2.0_rec
+    "korean_pp_rec_OCRv3": "rec/ppocr/korean_PP-OCRv3_rec.yaml",                     # korean_PP-OCRv3_rec
+    "japan_pp_rec_OCRv3": "rec/ppocr/japan_PP-OCRv3_rec.yaml",                       # japan_PP-OCRv3_rec
+    "chinese_cht_pp_rec_OCRv3": "rec/ppocr/chinese_cht_PP-OCRv3_rec.yaml",           # chinese_cht_PP-OCRv3_rec
+    "te_pp_rec_OCRv3": "rec/ppocr/te_PP-OCRv3_rec.yaml",                             # te_PP-OCRv3_rec
+    "ka_pp_rec_OCRv3": "rec/ppocr/ka_PP-OCRv3_rec.yaml",                             # ka_PP-OCRv3_rec
+    "ta_pp_rec_OCRv3": "rec/ppocr/ta_PP-OCRv3_rec.yaml",                             # ta_PP-OCRv3_rec
+    "latin_pp_rec_OCRv3": "rec/ppocr/latin_PP-OCRv3_rec.yaml",                       # latin_PP-OCRv3_rec
+    "arabic_pp_rec_OCRv3": "rec/ppocr/arabic_PP-OCRv3_rec.yaml",                     # arabic_PP-OCRv3_rec
+    "cyrillic_pp_rec_OCRv3": "rec/ppocr/cyrillic_PP-OCRv3_rec.yaml",                 # cyrillic_PP-OCRv3_rec
+    "devanagari_pp_rec_OCRv3": "rec/ppocr/devanagari_PP-OCRv3_rec.yaml"              # devanagari_PP-OCRv3_rec
+
 }