Skip to content

📄 Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.

Add pytest

SWHL334f0f13e6d7
diff --git a/python/README.md b/python/README.md
index 6e224a4..13026e8 100755
--- a/python/README.md
+++ b/python/README.md
@@ -89,6 +89,11 @@
- Note: 在Windows端,Shapely库可能自动安装会有问题,解决方案参见[Q15](../docs/FAQ.md#q15-装完环境之后运行python-mainpy之后报错oserror-winerror-126-找不到指定的模組)
4. 运行示例
+ - 运行单元测试
+ ```bash
+ cd tests
+ pytest test_*.py
+ ```
- 接口调用
```python
import cv2
@@ -129,7 +134,7 @@
|Global| `text_score` | [0, 1] | 0.5 | 文本识别结果置信度,值越大,把握越大 |
|Global| `use_angle_cls` | `bool` | `true` | 是否使用文本行的方向分类 |
|Global| `print_verbose` | `bool` | `true` | 是否打印各个部分耗时信息 |
-|Global| `min_heigh` | `int` | 50 | 图像高度,单位是像素 |
+|Global| `min_heigh` | `int` | 30 | 图像最小高度(单位是像素),低于这个值,会跳过文本检测阶段,直接进行后续识别 |
||||||
|Det| `box_thresh` | [0, 1] | 0.5 | 文本检测所得框是否保留的阈值,值越大,召回率越低 |
|Det| `unclip_ratio` | [1.6, 2.0] | 1.6 | 控制文本检测框的大小,值越大,检测框整体越大 |
diff --git a/python/rapidocr_onnxruntime/__init__.py b/python/rapidocr_onnxruntime/__init__.py
index 90121b5..e0e6353 100644
--- a/python/rapidocr_onnxruntime/__init__.py
+++ b/python/rapidocr_onnxruntime/__init__.py
@@ -1,4 +1,4 @@
# -*- encoding: utf-8 -*-
# @Author: SWHL
# @Contact: [email protected]
-from .rapid_ocr_api import TextSystem
\ No newline at end of file
+from .rapid_ocr_api import TextSystem
diff --git a/python/rapidocr_onnxruntime/ch_ppocr_v2_cls/text_cls.py b/python/rapidocr_onnxruntime/ch_ppocr_v2_cls/text_cls.py
index 38d8549..bdd6e17 100644
--- a/python/rapidocr_onnxruntime/ch_ppocr_v2_cls/text_cls.py
+++ b/python/rapidocr_onnxruntime/ch_ppocr_v2_cls/text_cls.py
@@ -19,7 +19,6 @@ from typing import List
import cv2
import numpy as np
-import onnxruntime as ort
try:
from .utils import ClsPostProcess, read_yaml, OrtInferSession
@@ -38,29 +37,6 @@ class TextClassifier(object):
self.session = session_instance.session
self.input_name = session_instance.get_input_name()
- def resize_norm_img(self, img):
- img_c, img_h, img_w = self.cls_image_shape
- h, w = img.shape[:2]
- ratio = w / float(h)
- if math.ceil(img_h * ratio) > img_w:
- resized_w = img_w
- else:
- resized_w = int(math.ceil(img_h * ratio))
-
- resized_image = cv2.resize(img, (resized_w, img_h))
- resized_image = resized_image.astype('float32')
- if img_c == 1:
- resized_image = resized_image / 255
- resized_image = resized_image[np.newaxis, :]
- else:
- resized_image = resized_image.transpose((2, 0, 1)) / 255
-
- resized_image -= 0.5
- resized_image /= 0.5
- padding_im = np.zeros((img_c, img_h, img_w), dtype=np.float32)
- padding_im[:, :, :resized_w] = resized_image
- return padding_im
-
def __call__(self, img_list: List[np.ndarray]):
if isinstance(img_list, np.ndarray):
img_list = [img_list]
@@ -106,6 +82,29 @@ class TextClassifier(object):
img_list[indices[beg_img_no + rno]], 1)
return img_list, cls_res, elapse
+ def resize_norm_img(self, img):
+ img_c, img_h, img_w = self.cls_image_shape
+ h, w = img.shape[:2]
+ ratio = w / float(h)
+ if math.ceil(img_h * ratio) > img_w:
+ resized_w = img_w
+ else:
+ resized_w = int(math.ceil(img_h * ratio))
+
+ resized_image = cv2.resize(img, (resized_w, img_h))
+ resized_image = resized_image.astype('float32')
+ if img_c == 1:
+ resized_image = resized_image / 255
+ resized_image = resized_image[np.newaxis, :]
+ else:
+ resized_image = resized_image.transpose((2, 0, 1)) / 255
+
+ resized_image -= 0.5
+ resized_image /= 0.5
+ padding_im = np.zeros((img_c, img_h, img_w), dtype=np.float32)
+ padding_im[:, :, :resized_w] = resized_image
+ return padding_im
+
if __name__ == "__main__":
parser = argparse.ArgumentParser()
diff --git a/python/rapidocr_onnxruntime/ch_ppocr_v2_det/config.yaml b/python/rapidocr_onnxruntime/ch_ppocr_v2_det/config.yaml
index 428ccfc..21bc186 100644
--- a/python/rapidocr_onnxruntime/ch_ppocr_v2_det/config.yaml
+++ b/python/rapidocr_onnxruntime/ch_ppocr_v2_det/config.yaml
@@ -1,11 +1,9 @@
-model_path: G:\ProgramFiles\_self\RapidOCR\python\resources\models\ch_PP-OCRv2_det_infer.onnx
+model_path: resources/models/ch_PP-OCRv2_det_infer.onnx
use_cuda: false
-# Details of the params: https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html
CUDAExecutionProvider:
device_id: 0
arena_extend_strategy: kNextPowerOfTwo
- gpu_mem_limit: 2 * 1024 * 1024 * 1024
cudnn_conv_algo_search: EXHAUSTIVE
do_copy_in_default_stream: true
diff --git a/python/rapidocr_onnxruntime/ch_ppocr_v2_det/text_detect.py b/python/rapidocr_onnxruntime/ch_ppocr_v2_det/text_detect.py
index 19b6b31..e04c2b0 100644
--- a/python/rapidocr_onnxruntime/ch_ppocr_v2_det/text_detect.py
+++ b/python/rapidocr_onnxruntime/ch_ppocr_v2_det/text_detect.py
@@ -20,7 +20,6 @@ import time
import cv2
import numpy as np
-
try:
from .utils import (DBPostProcess, create_operators,
draw_text_det_res, transform, read_yaml, OrtInferSession)
diff --git a/python/rapidocr_openvino/ch_ppocr_v2_cls/text_cls.py b/python/rapidocr_openvino/ch_ppocr_v2_cls/text_cls.py
index 91c5b5c..b056524 100644
--- a/python/rapidocr_openvino/ch_ppocr_v2_cls/text_cls.py
+++ b/python/rapidocr_openvino/ch_ppocr_v2_cls/text_cls.py
@@ -40,29 +40,6 @@ class TextClassifier(object):
self.session = ort.InferenceSession(config['model_path'], sess_opt)
self.input_name = self.session.get_inputs()[0].name
- def resize_norm_img(self, img):
- img_c, img_h, img_w = self.cls_image_shape
- h, w = img.shape[:2]
- ratio = w / float(h)
- if math.ceil(img_h * ratio) > img_w:
- resized_w = img_w
- else:
- resized_w = int(math.ceil(img_h * ratio))
-
- resized_image = cv2.resize(img, (resized_w, img_h))
- resized_image = resized_image.astype('float32')
- if img_c == 1:
- resized_image = resized_image / 255
- resized_image = resized_image[np.newaxis, :]
- else:
- resized_image = resized_image.transpose((2, 0, 1)) / 255
-
- resized_image -= 0.5
- resized_image /= 0.5
- padding_im = np.zeros((img_c, img_h, img_w), dtype=np.float32)
- padding_im[:, :, :resized_w] = resized_image
- return padding_im
-
def __call__(self, img_list: List[np.ndarray]):
if isinstance(img_list, np.ndarray):
img_list = [img_list]
@@ -108,6 +85,29 @@ class TextClassifier(object):
img_list[indices[beg_img_no + rno]], 1)
return img_list, cls_res, elapse
+ def resize_norm_img(self, img):
+ img_c, img_h, img_w = self.cls_image_shape
+ h, w = img.shape[:2]
+ ratio = w / float(h)
+ if math.ceil(img_h * ratio) > img_w:
+ resized_w = img_w
+ else:
+ resized_w = int(math.ceil(img_h * ratio))
+
+ resized_image = cv2.resize(img, (resized_w, img_h))
+ resized_image = resized_image.astype('float32')
+ if img_c == 1:
+ resized_image = resized_image / 255
+ resized_image = resized_image[np.newaxis, :]
+ else:
+ resized_image = resized_image.transpose((2, 0, 1)) / 255
+
+ resized_image -= 0.5
+ resized_image /= 0.5
+ padding_im = np.zeros((img_c, img_h, img_w), dtype=np.float32)
+ padding_im[:, :, :resized_w] = resized_image
+ return padding_im
+
if __name__ == "__main__":
parser = argparse.ArgumentParser()
diff --git a/python/requirements.txt b/python/requirements.txt
index fb2231c..ac08210 100644
--- a/python/requirements.txt
+++ b/python/requirements.txt
@@ -4,4 +4,5 @@ opencv_python==4.5.1.48
six>=1.15.0
numpy>=1.19.2
Pillow==8.3.0
-PyYAML
\ No newline at end of file
+PyYAML
+pytest
\ No newline at end of file
diff --git a/python/tests/base.py b/python/tests/base.py
new file mode 100644
index 0000000..8b07ee7
--- /dev/null
+++ b/python/tests/base.py
@@ -0,0 +1,26 @@
+# -*- encoding: utf-8 -*-
+# @Author: SWHL
+# @Contact: [email protected]
+
+import importlib
+from pathlib import Path
+import sys
+
+import yaml
+
+package_name = 'rapidocr_onnxruntime'
+
+root_dir = Path(__file__).resolve().parent.parent
+sys.path.append(str(root_dir))
+sys.path.append(str(root_dir / package_name))
+
+
+def init_module(module_name, class_name):
+ module_part = importlib.import_module(module_name)
+ return getattr(module_part, class_name)
+
+
+def read_yaml(yaml_path):
+ with open(yaml_path, 'rb') as f:
+ data = yaml.load(f, Loader=yaml.Loader)
+ return data
\ No newline at end of file
diff --git a/python/tests/test_cls.py b/python/tests/test_cls.py
new file mode 100644
index 0000000..d8c6dc5
--- /dev/null
+++ b/python/tests/test_cls.py
@@ -0,0 +1,23 @@
+# -*- encoding: utf-8 -*-
+# @Author: SWHL
+# @Contact: [email protected]
+import base
+import cv2
+
+
+module_name = 'ch_ppocr_v2_cls'
+class_name = 'TextClassifier'
+TextClassifier = base.init_module(module_name, class_name)
+
+yaml_path = base.root_dir / base.package_name / module_name / 'config.yaml'
+config = base.read_yaml(str(yaml_path))
+config['model_path'] = str(base.root_dir / config['model_path'])
+
+text_cls = TextClassifier(config)
+
+
+def test_cls():
+ img_path = 'test_files/text_cls.jpg'
+ img = cv2.imread(img_path)
+ result = text_cls([img])
+ assert result[1][0][0] == '180'
diff --git a/python/tests/test_det.py b/python/tests/test_det.py
new file mode 100644
index 0000000..b0f8f2a
--- /dev/null
+++ b/python/tests/test_det.py
@@ -0,0 +1,25 @@
+# -*- encoding: utf-8 -*-
+# @Author: SWHL
+# @Contact: [email protected]
+import base
+import cv2
+
+
+module_name = 'ch_ppocr_v2_det'
+class_name = 'TextDetector'
+TextDetector = base.init_module(module_name, class_name)
+
+yaml_path = base.root_dir / base.package_name / module_name / 'config.yaml'
+config = base.read_yaml(str(yaml_path))
+config['model_path'] = str(base.root_dir / config['model_path'])
+
+text_det = TextDetector(config)
+
+
+def test_det():
+ img_path = 'test_files/ch_en_num.jpg'
+
+ img = cv2.imread(img_path)
+
+ dt_boxes, elapse = text_det(img)
+ assert dt_boxes.shape == (18, 4, 2)
diff --git a/python/tests/test_files/ch_en_num.jpg b/python/tests/test_files/ch_en_num.jpg
new file mode 100644
index 0000000..d352416
Binary files /dev/null and b/python/tests/test_files/ch_en_num.jpg differ
diff --git a/python/tests/test_files/text_cls.jpg b/python/tests/test_files/text_cls.jpg
new file mode 100644
index 0000000..fcd7e12
Binary files /dev/null and b/python/tests/test_files/text_cls.jpg differ