clean code (#826)

Co-authored-by: qiuyufeng <[email protected]>

Author: QYuFong

mindocr/data/transforms/det_east_transforms.py

@@ -1,3 +1,5 @@
+import ast
+import json
 import math
 
 import cv2
@@ -414,8 +416,17 @@ def _extract_vertices(self, data_labels):
         """
         vertices_list = []
         labels_list = []
-        data_labels = eval(data_labels)
-        for data_label in data_labels:
+        try:
+            parsed_data = json.loads(data_labels)
+        except json.JSONDecodeError:
+            try:
+                parsed_data = ast.literal_eval(data_labels)
+            except (ValueError, SyntaxError) as e:
+                raise ValueError(f"Invalid data format: {str(e)}") from e
+
+        if not isinstance(parsed_data, list):
+            raise ValueError("Data labels should be a list")
+        for data_label in parsed_data:
             vertices = data_label["points"]
             vertices = [item for point in vertices for item in point]
             vertices_list.append(vertices)

mindocr/data/transforms/svtr_transform.py

@@ -546,10 +546,19 @@ def __init__(self, max_text_length, character_dict_path=None, use_space_char=Fal
 
         self.ctc_encode = CTCLabelEncodeForSVTR(max_text_length, character_dict_path, use_space_char, **kwargs)
         self.gtc_encode_type = gtc_encode
+        # Pls explicitly specify the supported gtc_encode classes and obtain the class objects through dictionaries.
+        supported_gtc_encode = {}
         if gtc_encode is None:
             self.gtc_encode = SARLabelEncodeForSVTR(max_text_length, character_dict_path, use_space_char, **kwargs)
         else:
-            self.gtc_encode = eval(gtc_encode)(max_text_length, character_dict_path, use_space_char, **kwargs)
+            # Mindocr currently does not have a module that requires a custom `gtc_encode` input parameter, and will not
+            # enter this branch at present. If it is supported later, please directly obtain the class reference through
+            # a specific dict, and do not use the `eval` function.
+            if gtc_encode not in supported_gtc_encode:
+                raise ValueError(f"Get unsupported gtc_encode {gtc_encode}")
+            self.gtc_encode = supported_gtc_encode[gtc_encode](
+                max_text_length, character_dict_path, use_space_char, **kwargs
+            )
 
     def __call__(self, data):
         data_ctc = copy.deepcopy(data)
@@ -925,7 +934,7 @@ def __init__(
         jitter_prob=0.4,
         blur_prob=0.4,
         hsv_aug_prob=0.4,
-        **kwargs
+        **kwargs,
     ):
         self.crop_prob = crop_prob
         self.reverse_prob = reverse_prob
@@ -973,7 +982,7 @@ def __init__(
         jitter_prob=0.4,
         blur_prob=0.4,
         hsv_aug_prob=0.4,
-        **kwargs
+        **kwargs,
     ):
         self.tia_prob = tia_prob
         self.bda = BaseDataAugmentation(crop_prob, reverse_prob, noise_prob, jitter_prob, blur_prob, hsv_aug_prob)
@@ -1078,7 +1087,7 @@ def __init__(
         character_dict_path=".mindocr/utils/dict/ch_dict.txt",
         padding=True,
         width_downsample_ratio=0.125,
-        **kwargs
+        **kwargs,
     ):
         self.image_shape = image_shape
         self.infer_mode = infer_mode

mindocr/data/transforms/transforms_factory.py

@@ -1,11 +1,8 @@
 """
 Create and run transformations from a config or predefined transformation pipeline
 """
-import logging
 from typing import Dict, List
 
-import numpy as np
-
 from .det_east_transforms import *
 from .det_transforms import *
 from .general_transforms import *
@@ -15,6 +12,63 @@
 from .svtr_transform import *
 from .table_transform import *
 
+SUPPORTED_TRANSFORMS = {
+    "EASTProcessTrain": EASTProcessTrain,
+    "DetLabelEncode": DetLabelEncode,
+    "BorderMap": BorderMap,
+    "ShrinkBinaryMap": ShrinkBinaryMap,
+    "expand_poly": expand_poly,
+    "PSEGtDecode": PSEGtDecode,
+    "ValidatePolygons": ValidatePolygons,
+    "RandomCropWithBBox": RandomCropWithBBox,
+    "RandomCropWithMask": RandomCropWithMask,
+    "DetResize": DetResize,
+    "DecodeImage": DecodeImage,
+    "NormalizeImage": NormalizeImage,
+    "ToCHWImage": ToCHWImage,
+    "PackLoaderInputs": PackLoaderInputs,
+    "RandomScale": RandomScale,
+    "RandomColorAdjust": RandomColorAdjust,
+    "RandomRotate": RandomRotate,
+    "RandomHorizontalFlip": RandomHorizontalFlip,
+    "LayoutResize": LayoutResize,
+    "ImageStridePad": ImageStridePad,
+    "VQATokenLabelEncode": VQATokenLabelEncode,
+    "VQATokenPad": VQATokenPad,
+    "VQASerTokenChunk": VQASerTokenChunk,
+    "VQAReTokenRelation": VQAReTokenRelation,
+    "VQAReTokenChunk": VQAReTokenChunk,
+    "TensorizeEntitiesRelations": TensorizeEntitiesRelations,
+    "ABINetTransforms": ABINetTransforms,
+    "ABINetRecAug": ABINetRecAug,
+    "ABINetEval": ABINetEval,
+    "ABINetEvalTransforms": ABINetEvalTransforms,
+    "RecCTCLabelEncode": RecCTCLabelEncode,
+    "RecAttnLabelEncode": RecAttnLabelEncode,
+    "RecMasterLabelEncode": RecMasterLabelEncode,
+    "VisionLANLabelEncode": VisionLANLabelEncode,
+    "RecResizeImg": RecResizeImg,
+    "RecResizeNormForInfer": RecResizeNormForInfer,
+    "SVTRRecResizeImg": SVTRRecResizeImg,
+    "Rotate90IfVertical": Rotate90IfVertical,
+    "ClsLabelEncode": ClsLabelEncode,
+    "SARLabelEncode": SARLabelEncode,
+    "RobustScannerRecResizeImg": RobustScannerRecResizeImg,
+    "SVTRRecAug": SVTRRecAug,
+    "MultiLabelEncode": MultiLabelEncode,
+    "RecConAug": RecConAug,
+    "RecAug": RecAug,
+    "RecResizeImgForSVTR": RecResizeImgForSVTR,
+    "BaseRecLabelEncode": BaseRecLabelEncode,
+    "AttnLabelEncode": AttnLabelEncode,
+    "TableLabelEncode": TableLabelEncode,
+    "TableMasterLabelEncode": TableMasterLabelEncode,
+    "ResizeTableImage": ResizeTableImage,
+    "PaddingTableImage": PaddingTableImage,
+    "TableBoxEncode": TableBoxEncode,
+    "TableImageNorm": TableImageNorm,
+}
+
 __all__ = ["create_transforms", "run_transforms", "transforms_dbnet_icdar15"]
 _logger = logging.getLogger(__name__)
 
@@ -45,9 +99,9 @@ def create_transforms(transform_pipeline: List, global_config: Dict = None):
             param = {} if transform_config[trans_name] is None else transform_config[trans_name]
             if global_config is not None:
                 param.update(global_config)
-            # TODO: assert undefined transform class
-
-            transform = eval(trans_name)(**param)
+            # For security reasons, we no longer use the eval function to dynamically obtain class objects.
+            # If you need to add a new transform class, please explicitly add it to the ``SUPPORTED_TRANSFORMS`` dict.
+            transform = SUPPORTED_TRANSFORMS[trans_name](**param)
             transforms.append(transform)
         elif callable(transform_config):
             transforms.append(transform_config)

mindocr/models/backbones/mindcv_models/download.py

@@ -99,10 +99,11 @@ def extract_archive(self, from_path: str, to_path: str = None) -> str:
     def download_file(self, url: str, file_path: str, chunk_size: int = 1024):
         """Download a file."""
 
-        # no check certificate
+        # For security reasons, this repository code does not provide a function to disable SSL.
+        # If necessary, please disable SSL verification yourself.
         ctx = ssl.create_default_context()
-        ctx.check_hostname = False
-        ctx.verify_mode = ssl.CERT_NONE
+        # ctx.check_hostname = False
+        # ctx.verify_mode = ssl.CERT_NONE
 
         # Define request headers.
         headers = {"User-Agent": self.USER_AGENT}

mindocr/postprocess/builder.py

@@ -23,18 +23,24 @@
 
 __all__ = ["build_postprocess"]
 
-supported_postprocess = (
-    det_db_postprocess.__all__
-    + det_pse_postprocess.__all__
-    + det_east_postprocess.__all__
-    + rec_postprocess.__all__
-    + cls_postprocess.__all__
-    + rec_abinet_postprocess.__all__
-    + kie_ser_postprocess.__all__
-    + kie_re_postprocess.__all__
-    + layout_postprocess.__all__
-    + table_postprocess.__all__
-)
+SUPPORTED_POSTPROCESS = {
+    "DBPostprocess": DBPostprocess,
+    "PSEPostprocess": PSEPostprocess,
+    "EASTPostprocess": EASTPostprocess,
+    "CTCLabelDecode": CTCLabelDecode,
+    "RecCTCLabelDecode": RecCTCLabelDecode,
+    "RecAttnLabelDecode": RecAttnLabelDecode,
+    "RecMasterLabelDecode": RecMasterLabelDecode,
+    "VisionLANPostProcess": VisionLANPostProcess,
+    "SARLabelDecode": SARLabelDecode,
+    "ClsPostprocess": ClsPostprocess,
+    "ABINetLabelDecode": ABINetLabelDecode,
+    "VQASerTokenLayoutLMPostProcess": VQASerTokenLayoutLMPostProcess,
+    "VQAReTokenLayoutLMPostProcess": VQAReTokenLayoutLMPostProcess,
+    "YOLOv8Postprocess": YOLOv8Postprocess,
+    "Layoutlmv3Postprocess": Layoutlmv3Postprocess,
+    "TableMasterLabelDecode": TableMasterLabelDecode,
+}
 
 
 def build_postprocess(config: dict):
@@ -57,11 +63,11 @@ def build_postprocess(config: dict):
         >>> postprocess
     """
     proc = config.pop("name")
-    if proc in supported_postprocess:
-        postprocessor = eval(proc)(**config)
-    elif proc is None:
+    if proc is None:
         return None
+    if proc in SUPPORTED_POSTPROCESS:
+        postprocessor = SUPPORTED_POSTPROCESS[proc](**config)
     else:
-        raise ValueError(f"Invalid postprocess name {proc}, support postprocess are {supported_postprocess}")
+        raise ValueError(f"Invalid postprocess name {proc}, support postprocess are {SUPPORTED_POSTPROCESS.keys()}")
 
     return postprocessor

tools/arg_parser.py

@@ -62,7 +62,10 @@ def _parse_options(opts: list):
             "=" in opt_str
         ), "Invalid option {}. A valid option must be in the format of {{key_name}}={{value}}".format(opt_str)
         k, v = opt_str.strip().split("=")
-        options[k] = yaml.load(v, Loader=yaml.Loader)
+        try:
+            options[k] = yaml.load(v, Loader=yaml.SafeLoader)
+        except yaml.YAMLError as e:
+            raise ValueError(f"Failed to parse value for key '{k}': {str(e)}") from e
     # print('Parsed options: ', options)
 
     return options