Skip to content

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

修复路径问题,调大一些检测框

SWHL32a17de97fed
diff --git a/ocrweb/main.py b/ocrweb/main.py
index 308f57c..90e19be 100644
--- a/ocrweb/main.py
+++ b/ocrweb/main.py
@@ -12,9 +12,9 @@ from werkzeug.utils import secure_filename
from resources.bpocr import TextSystem, draw_text_det_res
# 实例化模型
-det_model_path = 'resources/models/ch_ppocr_mobile_v2_det_train.onnx'
+det_model_path = 'resources/models/ch_ppocr_mobile_v2.0_det_infer.onnx'
cls_model_path = 'resources/models/ch_ppocr_mobile_v2.0_cls_infer.onnx'
-rec_model_path = 'resources/models/ch_ppocr_mobile_v2.0_rec_pre_infer.onnx'
+rec_model_path = 'resources/models/ch_ppocr_mobile_v2.0_rec_infer.onnx'
text_sys = TextSystem(det_model_path,
rec_model_path,
@@ -24,16 +24,25 @@ text_sys = TextSystem(det_model_path,
# 设置允许的文件格式
ALLOWED_EXTENSIONS = {'png', 'jpg', 'JPG', 'PNG', 'bmp', 'jpeg', 'gif', 'tif'}
app = Flask(__name__)
+
+# 设置上传文件大小
+app.config['MAX_CONTENT_LENGTH'] = 10 * 1024 * 1024 # 10M
+
# 设置静态文件缓存过期时间
# app.send_file_max_age_default = TimedeltaFormat(seconds=10)
+def make_directory(directory_path):
+ if not isinstance(directory_path, Path):
+ directory_path = Path(directory_path)
+ directory_path.mkdir(parents=True, exist_ok=True)
+
def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1] in ALLOWED_EXTENSIONS
[email protected]('/upload', methods=['POST', 'GET'])
-def upload():
[email protected]('/ocr', methods=['POST', 'GET'])
+def ocr():
if request.method == 'POST':
f = request.files['file']
@@ -46,8 +55,15 @@ def upload():
img_name = f"{time_stamp}_{secure_filename(f.filename)}"
root_path = Path(__file__).resolve().parent
- static_path = root_path / 'static'
- upload_path = str(static_path / 'images' / f'raw_{img_name}')
+ static_directory = root_path / 'static'
+
+ raw_directory = static_directory / 'images'
+ make_directory(raw_directory)
+
+ infer_directory = static_directory / 'inference_results'
+ make_directory(infer_directory)
+
+ upload_path = str(raw_directory / f'raw_{img_name}')
f.save(upload_path)
t1 = time.time()
@@ -56,13 +72,13 @@ def upload():
rec_res = list(zip(range(len(rec_res)), rec_res))
- plot_im = draw_text_det_res(dt_boxes, img)
- plot_path = f'inference_results/det_{img_name}'
+ det_im = draw_text_det_res(dt_boxes, img)
+ det_path = f'inference_results/det_{img_name}'
- cv2.imwrite(str(static_path / plot_path), plot_im)
+ cv2.imwrite(str(static_directory / det_path), det_im)
return render_template('index_ok.html',
- det_img_path=plot_path,
+ det_img_path=det_path,
rec_res=rec_res,
time=str(f'{t2-t1:.3f}'),
width='100%',
@@ -72,4 +88,4 @@ def upload():
if __name__ == '__main__':
- app.run(host='0.0.0.0', port=9003, debug=True)
+ app.run(host='0.0.0.0', port=9003, debug=False, processes=True)
diff --git a/ocrweb/readme.md b/ocrweb/readme.md
index 446baab..541d1f0 100755
--- a/ocrweb/readme.md
+++ b/ocrweb/readme.md
@@ -1 +1,3 @@
-The whole project for web based OCR application.
+#### The whole project for web based OCR application.
+
+#### TODO
diff --git a/ocrweb/resources/bpocr.py b/ocrweb/resources/bpocr.py
index 1a86a24..489bb51 100644
--- a/ocrweb/resources/bpocr.py
+++ b/ocrweb/resources/bpocr.py
@@ -33,7 +33,7 @@ def draw_text_det_res(dt_boxes, raw_im):
box = np.array(box).astype(np.int32).reshape(-1, 2)
cv2.polylines(src_im, [box], True,
color=(0, 0, 255),
- thickness=2)
+ thickness=1)
cv2.putText(src_im, str(i), (int(box[0][0]), int(box[0][1])),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0), 2)
return src_im
diff --git a/ocrweb/resources/ch_ppocr_mobile_v2_det/text_detect.py b/ocrweb/resources/ch_ppocr_mobile_v2_det/text_detect.py
index 0bf6163..78abb82 100644
--- a/ocrweb/resources/ch_ppocr_mobile_v2_det/text_detect.py
+++ b/ocrweb/resources/ch_ppocr_mobile_v2_det/text_detect.py
@@ -54,7 +54,7 @@ class TextDetector(object):
self.postprocess_op = DBPostProcess(thresh=0.3,
box_thresh=0.5,
max_candidates=1000,
- unclip_ratio=1.6,
+ unclip_ratio=2.0,
use_dilation=True)
self.session = onnxruntime.InferenceSession(det_model_path)
self.input_name = self.session.get_inputs()[0].name
diff --git a/ocrweb/templates/index_ok.html b/ocrweb/templates/index_ok.html
index f940544..e4a017c 100644
--- a/ocrweb/templates/index_ok.html
+++ b/ocrweb/templates/index_ok.html
@@ -6,8 +6,8 @@
</head>
<style type="text/css">
body{
- background-color:#ebedef;
-}
+ background-color:#ebedef;
+ }
.area{
text-align: center;
height: auto;
@@ -27,16 +27,9 @@
.table{
margin: 0 auto;
+ border: 0;
/* border-spacing: 0px 15px; */
}
- tr td{
- border:1px solid black;
- border-bottom-width: thin;
- border-right-style: thin;
- border-left-style: thin;
- border-top-style: thin;
- }
-
.btn {
margin-top: 4px;
line-height: 32px;
@@ -90,7 +83,7 @@
<div class="rightarea">
<span class="span_title">文本识别结果(耗时:{{time}}s):</span>
<br/> <br/>
- <table class='table' cellpadding="0" cellspacing="0">
+ <table class="table" cellpadding="0" cellspacing="0">
<tr>
<td><strong>序号</strong></td>
<td><strong>识别结果</strong></td>