Spaces:
Running
Running
add back loadimg
Browse files
app.py
CHANGED
|
@@ -47,18 +47,19 @@ model.config.pad_token_id = tokenizer.eos_token_id
|
|
| 47 |
|
| 48 |
@spaces.GPU
|
| 49 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None, render=False):
|
|
|
|
| 50 |
if task == "Plain Text OCR":
|
| 51 |
-
res = model.chat(tokenizer,
|
| 52 |
elif task == "Format Text OCR":
|
| 53 |
-
res = model.chat(tokenizer,
|
| 54 |
elif task == "Fine-grained OCR (Box)":
|
| 55 |
-
res = model.chat(tokenizer,
|
| 56 |
elif task == "Fine-grained OCR (Color)":
|
| 57 |
-
res = model.chat(tokenizer,
|
| 58 |
elif task == "Multi-crop OCR":
|
| 59 |
-
res = model.chat_crop(tokenizer, image_file=
|
| 60 |
elif task == "Render Formatted OCR":
|
| 61 |
-
res = model.chat(tokenizer,
|
| 62 |
with open('./results/demo.html', 'r') as f:
|
| 63 |
html_content = f.read()
|
| 64 |
return res, html_content
|
|
@@ -157,4 +158,5 @@ with gr.Blocks() as demo:
|
|
| 157 |
outputs=[output_text, output_html]
|
| 158 |
)
|
| 159 |
|
| 160 |
-
|
|
|
|
|
|
| 47 |
|
| 48 |
@spaces.GPU
|
| 49 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None, render=False):
|
| 50 |
+
img = load_img(image, output_type="pil", input_type="auto")
|
| 51 |
if task == "Plain Text OCR":
|
| 52 |
+
res = model.chat(tokenizer, img, ocr_type='ocr')
|
| 53 |
elif task == "Format Text OCR":
|
| 54 |
+
res = model.chat(tokenizer, img, ocr_type='format')
|
| 55 |
elif task == "Fine-grained OCR (Box)":
|
| 56 |
+
res = model.chat(tokenizer, img, ocr_type=ocr_type, ocr_box=ocr_box)
|
| 57 |
elif task == "Fine-grained OCR (Color)":
|
| 58 |
+
res = model.chat(tokenizer, img, ocr_type=ocr_type, ocr_color=ocr_color)
|
| 59 |
elif task == "Multi-crop OCR":
|
| 60 |
+
res = model.chat_crop(tokenizer, image_file=img)
|
| 61 |
elif task == "Render Formatted OCR":
|
| 62 |
+
res = model.chat(tokenizer, img, ocr_type='format', render=True, save_render_file='./results/demo.html')
|
| 63 |
with open('./results/demo.html', 'r') as f:
|
| 64 |
html_content = f.read()
|
| 65 |
return res, html_content
|
|
|
|
| 158 |
outputs=[output_text, output_html]
|
| 159 |
)
|
| 160 |
|
| 161 |
+
if __name__ == "__main__":
|
| 162 |
+
demo.launch()
|