debug

app.py

@@ -162,23 +162,17 @@ def apply_half(t):
     return t
 
 
-def ocr(img, task_type):
+def ocr(img):
     out_img = Image.open(img)
-    if task_type == "General":
-        results = get_images(img, reader, text_confidence=0.7, text_threshold=0.43,
-                             link_threshold=0.43, slope_ths=0., add_margin=0.02)
-    elif task_type == "Document":
-        results = get_images(img, reader, text_threshold=0.3, sharp=1.2,
-                             contrast=1.2, space_ths=1.8)
-    else:
-        raise NotImplementedError
+    results = get_images(img, reader, text_confidence=0.7, text_threshold=0.43,
+                         link_threshold=0.43, slope_ths=0., add_margin=0.02)
     box_list, image_list = zip(*results)
     draw_boxes(out_img, box_list)
 
     ocr_result = []
     for i, (box, image) in enumerate(zip(box_list, image_list)):
         image = Image.fromarray(image)
-        sample = construct_sample(image, cfg.task.patch_image_size, is_document=False)
+        sample = construct_sample(image, cfg.task.patch_image_size)
         sample = utils.move_to_cuda(sample) if use_cuda else sample
         sample = utils.apply_to_sample(apply_half, sample) if use_fp16 else sample
 
@@ -198,11 +192,10 @@ description = "Gradio Demo for Chinese OCR based on OFA-Base. "\
               "\n中文OCR体验区。欢迎上传图片，静待检测文字返回~"
 article = "<p style='text-align: center'><a href='https://github.com/OFA-Sys/OFA' target='_blank'>OFA Github " \
           "Repo</a></p> "
-examples = [['shupai.png', 'General'], ['chinese.jpg', 'General'], ['gaidao.jpeg', 'General'],
-            ['qiaodaima.png', 'General'], ['benpao.jpeg', 'General'], ['wanli.png', 'General'],
+examples = [['shupai.png'], ['chinese.jpg'], ['gaidao.jpeg'],
+            ['qiaodaima.png'], ['benpao.jpeg'], ['wanli.png'],
             ['xsd.jpg', 'General']]
-io = gr.Interface(fn=ocr, inputs=[gr.inputs.Image(type='filepath', label='Image'),
-                  gr.inputs.Radio(choices=['General', 'Document'], type='value', default='General', label='Domain')],
+io = gr.Interface(fn=ocr, inputs=gr.inputs.Image(type='filepath', label='Image'),
                   outputs=[gr.outputs.Image(type='pil', label='Image'),
                            gr.outputs.Dataframe(headers=['Box ID', 'Text'], type='pandas', label='OCR Results')],
                   title=title, description=description, article=article, examples=examples)