chore: update files

python/rapidocr/ch_ppocr_cls/main.py

@@ -19,7 +19,7 @@
 import cv2
 import numpy as np
 
-from rapidocr.utils import OrtInferSession
+from rapidocr.inference_engine import OrtInferSession
 
 from .utils import ClsPostProcess, TextClsOutput
 

python/rapidocr/ch_ppocr_det/main.py

@@ -19,7 +19,7 @@
 
 import numpy as np
 
-from rapidocr.utils import OrtInferSession
+from rapidocr.inference_engine import OrtInferSession
 
 from .utils import DBPostProcess, DetPreProcess, TextDetOutput
 

python/rapidocr/ch_ppocr_rec/main.py

@@ -18,7 +18,7 @@
 import cv2
 import numpy as np
 
-from rapidocr.utils import OrtInferSession
+from rapidocr.inference_engine import OrtInferSession
 
 from .utils import CTCLabelDecode, TextRecInput, TextRecOutput
 

python/rapidocr/inference_engine/__init__.py

@@ -1,3 +1,4 @@
 # -*- encoding: utf-8 -*-
 # @Author: SWHL
 # @Contact: [email protected]
+from .onnxruntime import OrtInferSession

python/rapidocr/inference_engine/base.py

@@ -0,0 +1,30 @@
+# -*- encoding: utf-8 -*-
+# @Author: SWHL
+# @Contact: [email protected]
+import abc
+from pathlib import Path
+from typing import Union
+
+import numpy as np
+
+
+class InferSession(abc.ABC):
+    @abc.abstractmethod
+    def __init__(self, config):
+        pass
+
+    @abc.abstractmethod
+    def __call__(self, input_content: np.ndarray) -> np.ndarray:
+        pass
+
+    @staticmethod
+    def _verify_model(model_path: Union[str, Path, None]):
+        if model_path is None:
+            raise ValueError("model_path is None!")
+
+        model_path = Path(model_path)
+        if not model_path.exists():
+            raise FileNotFoundError(f"{model_path} does not exists.")
+
+        if not model_path.is_file():
+            raise FileExistsError(f"{model_path} is not a file.")

python/rapidocr/inference_engine/onnxruntime.py

@@ -5,8 +5,7 @@
 import platform
 import traceback
 from enum import Enum
-from pathlib import Path
-from typing import Any, Dict, List, Tuple, Union
+from typing import Any, Dict, List, Tuple
 
 import numpy as np
 from onnxruntime import (
@@ -18,6 +17,7 @@
 )
 
 from ..utils.logger import get_logger
+from .base import InferSession
 
 
 class EP(Enum):
@@ -26,7 +26,7 @@ class EP(Enum):
     DIRECTML_EP = "DmlExecutionProvider"
 
 
-class OrtInferSession:
+class OrtInferSession(InferSession):
     def __init__(self, config: Dict[str, Any]):
         self.logger = get_logger("OrtInferSession")
 
@@ -214,18 +214,6 @@ def have_key(self, key: str = "character") -> bool:
             return True
         return False
 
-    @staticmethod
-    def _verify_model(model_path: Union[str, Path, None]):
-        if model_path is None:
-            raise ValueError("model_path is None!")
-
-        model_path = Path(model_path)
-        if not model_path.exists():
-            raise FileNotFoundError(f"{model_path} does not exists.")
-
-        if not model_path.is_file():
-            raise FileExistsError(f"{model_path} is not a file.")
-
 
 class ONNXRuntimeError(Exception):
     pass

python/rapidocr/inference_engine/torch.py

@@ -1,3 +1,65 @@
 # -*- encoding: utf-8 -*-
 # @Author: SWHL
 # @Contact: [email protected]
+from pathlib import Path
+from typing import Dict, Optional, Union
+
+import numpy as np
+import torch
+import yaml
+
+root_dir = Path(__file__).resolve().parent.parent
+DEFAULT_CFG_PATH = root_dir / "arch_config.yaml"
+
+
+def read_yaml(yaml_path: Union[str, Path]) -> Dict[str, Dict]:
+    with open(yaml_path, "rb") as f:
+        data = yaml.load(f, Loader=yaml.Loader)
+    return data
+
+
+from rapidocr_torch.modeling.architectures.base_model import BaseModel
+
+from ..utils.logger import get_logger
+
+
+class TorchInferSession:
+    def __init__(self, config, mode: Optional[str] = None) -> None:
+        all_arch_config = read_yaml(DEFAULT_CFG_PATH)
+
+        self.logger = get_logger("TorchInferSession")
+        self.mode = mode
+        model_path = Path(config["model_path"])
+        self._verify_model(model_path)
+        file_name = model_path.stem
+        if file_name not in all_arch_config:
+            raise ValueError(f"architecture {file_name} is not in config.yaml")
+        arch_config = all_arch_config[file_name]
+        self.predictor = BaseModel(arch_config)
+        self.predictor.load_state_dict(torch.load(model_path, weights_only=True))
+        self.predictor.eval()
+        self.use_gpu = False
+        if config["use_cuda"]:
+            self.predictor.cuda()
+            self.use_gpu = True
+
+    def __call__(self, img: np.ndarray):
+        with torch.no_grad():
+            inp = torch.from_numpy(img)
+            if self.use_gpu:
+                inp = inp.cuda()
+            # 适配跟onnx对齐取值逻辑
+            outputs = self.predictor(inp).unsqueeze(0)
+            return outputs.cpu().numpy()
+
+    @staticmethod
+    def _verify_model(model_path):
+        model_path = Path(model_path)
+        if not model_path.exists():
+            raise FileNotFoundError(f"{model_path} does not exists.")
+        if not model_path.is_file():
+            raise FileExistsError(f"{model_path} is not a file.")
+
+
+class TorchInferError(Exception):
+    pass

python/rapidocr/main.py

@@ -44,8 +44,8 @@ def __init__(self, config_path: Optional[str] = None, **kwargs):
             updater = UpdateParameters()
             config = updater(config, **kwargs)
 
-        det_lang, rec_lang = parse_lang(config.Global.lang)
         # 根据选定的语言加载对应的模型
+        det_lang, rec_lang = parse_lang(config.Global.lang)
 
         self.print_verbose = config.Global.print_verbose
         self.text_score = config.Global.text_score

python/rapidocr/utils/__init__.py

@@ -1,7 +1,6 @@
 # -*- encoding: utf-8 -*-
 # @Author: SWHL
 # @Contact: [email protected]
-from ..inference_engine.onnxruntime import OrtInferSession
 from .load_image import LoadImage, LoadImageError
 from .logger import get_logger
 from .parse_parameters import UpdateParameters, init_args, parse_lang, update_model_path