Spaces:
Running
on
Zero
Running
on
Zero
Revert "initial commit"
Browse filesThis reverts commit b513b33c3b6bacce24ee5fba556a8ae7cfdf30d2.
app.py
CHANGED
@@ -44,21 +44,20 @@ model = AutoModel.from_pretrained('ucaslcl/GOT-OCR2_0', trust_remote_code=True,
|
|
44 |
model = model.eval().cuda()
|
45 |
model.config.pad_token_id = tokenizer.eos_token_id
|
46 |
|
47 |
-
|
48 |
@spaces.GPU
|
49 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None, render=True):
|
50 |
if task == "Plain Text OCR":
|
51 |
-
res = model.chat(tokenizer, image, ocr_type='ocr')
|
52 |
elif task == "Format Text OCR":
|
53 |
-
res = model.chat(tokenizer, image, ocr_type='format')
|
54 |
elif task == "Fine-grained OCR (Box)":
|
55 |
-
res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_box=ocr_box)
|
56 |
elif task == "Fine-grained OCR (Color)":
|
57 |
-
res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_color=ocr_color)
|
58 |
elif task == "Multi-crop OCR":
|
59 |
-
res = model.chat_crop(tokenizer, image_file=image)
|
60 |
elif task == "Render Formatted OCR":
|
61 |
-
res = model.chat(tokenizer, image, ocr_type='format', render=True, save_render_file='./demo.html')
|
62 |
with open('./demo.html', 'r') as f:
|
63 |
html_content = f.read()
|
64 |
return res, html_content
|
|
|
44 |
model = model.eval().cuda()
|
45 |
model.config.pad_token_id = tokenizer.eos_token_id
|
46 |
|
|
|
47 |
@spaces.GPU
|
48 |
def process_image(image, task, ocr_type=None, ocr_box=None, ocr_color=None, render=True):
|
49 |
if task == "Plain Text OCR":
|
50 |
+
res = model.chat(tokenizer, image, ocr_type='ocr', attention_mask=attention_mask)
|
51 |
elif task == "Format Text OCR":
|
52 |
+
res = model.chat(tokenizer, image, ocr_type='format', attention_mask=attention_mask)
|
53 |
elif task == "Fine-grained OCR (Box)":
|
54 |
+
res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_box=ocr_box, attention_mask=attention_mask)
|
55 |
elif task == "Fine-grained OCR (Color)":
|
56 |
+
res = model.chat(tokenizer, image, ocr_type=ocr_type, ocr_color=ocr_color, attention_mask=attention_mask)
|
57 |
elif task == "Multi-crop OCR":
|
58 |
+
res = model.chat_crop(tokenizer, image_file=image, attention_mask=attention_mask)
|
59 |
elif task == "Render Formatted OCR":
|
60 |
+
res = model.chat(tokenizer, image, ocr_type='format', render=True, save_render_file='./demo.html', attention_mask=attention_mask)
|
61 |
with open('./demo.html', 'r') as f:
|
62 |
html_content = f.read()
|
63 |
return res, html_content
|