Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -6,28 +6,29 @@ import traceback
|
|
6 |
from io import BytesIO
|
7 |
from typing import Any, Dict, List, Optional, Tuple
|
8 |
import re
|
9 |
-
from threading import Thread
|
10 |
import time
|
|
|
11 |
|
12 |
import gradio as gr
|
13 |
import requests
|
14 |
import torch
|
15 |
-
from PIL import Image
|
16 |
from transformers import (
|
17 |
Qwen2_5_VLForConditionalGeneration,
|
18 |
AutoProcessor,
|
19 |
TextIteratorStreamer,
|
20 |
)
|
|
|
21 |
|
22 |
# Constants
|
23 |
MIN_PIXELS = 3136
|
24 |
MAX_PIXELS = 11289600
|
25 |
IMAGE_FACTOR = 28
|
26 |
-
MAX_INPUT_TOKEN_LENGTH =
|
27 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
28 |
|
29 |
-
#
|
30 |
-
prompt = """Please output the layout information from the
|
31 |
|
32 |
1. Bbox format: [x1, y1, x2, y2]
|
33 |
|
@@ -46,74 +47,145 @@ prompt = """Please output the layout information from the PDF image, including e
|
|
46 |
5. Final Output: The entire output must be a single JSON object.
|
47 |
"""
|
48 |
|
49 |
-
# Load
|
50 |
MODEL_ID_M = "prithivMLmods/Camel-Doc-OCR-062825"
|
51 |
processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
|
52 |
model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
53 |
-
MODEL_ID_M,
|
|
|
|
|
54 |
).to(device).eval()
|
55 |
|
56 |
MODEL_ID_T = "prithivMLmods/Megalodon-OCR-Sync-0713"
|
57 |
processor_t = AutoProcessor.from_pretrained(MODEL_ID_T, trust_remote_code=True)
|
58 |
model_t = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
59 |
-
MODEL_ID_T,
|
|
|
|
|
60 |
).to(device).eval()
|
61 |
|
62 |
MODEL_ID_C = "nanonets/Nanonets-OCR-s"
|
63 |
processor_c = AutoProcessor.from_pretrained(MODEL_ID_C, trust_remote_code=True)
|
64 |
model_c = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
65 |
-
MODEL_ID_C,
|
|
|
|
|
66 |
).to(device).eval()
|
67 |
|
68 |
MODEL_ID_G = "echo840/MonkeyOCR"
|
69 |
SUBFOLDER = "Recognition"
|
70 |
processor_g = AutoProcessor.from_pretrained(
|
71 |
-
MODEL_ID_G,
|
|
|
|
|
72 |
)
|
73 |
model_g = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
74 |
-
MODEL_ID_G,
|
|
|
|
|
|
|
75 |
).to(device).eval()
|
76 |
|
77 |
-
|
78 |
# Utility functions
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
79 |
def is_arabic_text(text: str) -> bool:
|
80 |
-
"""Check if text contains mostly Arabic characters."""
|
81 |
if not text:
|
82 |
return False
|
83 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
arabic_chars = 0
|
85 |
total_chars = 0
|
86 |
-
for char in
|
87 |
if char.isalpha():
|
88 |
total_chars += 1
|
89 |
-
if '\u0600' <= char <= '\u06FF':
|
90 |
arabic_chars += 1
|
91 |
return total_chars > 0 and (arabic_chars / total_chars) > 0.5
|
92 |
|
93 |
def layoutjson2md(image: Image.Image, layout_data: List[Dict], text_key: str = 'text') -> str:
|
94 |
-
"""Convert layout JSON to markdown format."""
|
95 |
import base64
|
96 |
from io import BytesIO
|
97 |
markdown_lines = []
|
98 |
try:
|
99 |
-
# Sort items by reading order (top to bottom, left to right)
|
100 |
sorted_items = sorted(layout_data, key=lambda x: (x.get('bbox', [0, 0, 0, 0])[1], x.get('bbox', [0, 0, 0, 0])[0]))
|
101 |
for item in sorted_items:
|
102 |
category = item.get('category', '')
|
103 |
text = item.get(text_key, '')
|
104 |
bbox = item.get('bbox', [])
|
105 |
-
|
106 |
if category == 'Picture':
|
107 |
if bbox and len(bbox) == 4:
|
108 |
try:
|
109 |
-
x1, y1, x2, y2 =
|
110 |
-
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
|
|
|
|
|
|
|
|
|
|
115 |
except Exception as e:
|
116 |
-
|
|
|
|
|
|
|
117 |
elif not text:
|
118 |
continue
|
119 |
elif category == 'Title':
|
@@ -124,105 +196,161 @@ def layoutjson2md(image: Image.Image, layout_data: List[Dict], text_key: str = '
|
|
124 |
markdown_lines.append(f"{text}\n")
|
125 |
elif category == 'List-item':
|
126 |
markdown_lines.append(f"- {text}\n")
|
127 |
-
elif category == 'Table'
|
128 |
-
|
129 |
-
|
130 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
elif category == 'Caption':
|
132 |
markdown_lines.append(f"*{text}*\n")
|
133 |
elif category == 'Footnote':
|
134 |
-
|
135 |
-
elif category
|
|
|
|
|
136 |
markdown_lines.append(f"{text}\n")
|
|
|
137 |
except Exception as e:
|
138 |
print(f"Error converting to markdown: {e}")
|
139 |
-
return
|
140 |
return "\n".join(markdown_lines)
|
141 |
|
142 |
-
|
143 |
@spaces.GPU
|
144 |
-
def
|
145 |
-
"""
|
146 |
-
Generates a response using streaming, then processes the final output.
|
147 |
-
Yields updates for the raw stream, final markdown, and JSON output.
|
148 |
-
"""
|
149 |
-
if image is None:
|
150 |
-
yield "Please upload an image.", "Please upload an image.", None
|
151 |
-
return
|
152 |
-
|
153 |
-
# 1. Select Model and Processor
|
154 |
-
if model_name == "Camel-Doc-OCR-062825":
|
155 |
-
processor, model = processor_m, model_m
|
156 |
-
elif model_name == "Megalodon-OCR-Sync-0713":
|
157 |
-
processor, model = processor_t, model_t
|
158 |
-
elif model_name == "Nanonets-OCR-s":
|
159 |
-
processor, model = processor_c, model_c
|
160 |
-
elif model_name == "MonkeyOCR-Recognition":
|
161 |
-
processor, model = processor_g, model_g
|
162 |
-
else:
|
163 |
-
yield "Invalid model selected.", "Invalid model selected.", None
|
164 |
-
return
|
165 |
-
|
166 |
-
# 2. Prepare inputs for the model
|
167 |
-
messages = [{"role": "user", "content": [{"type": "image"}, {"type": "text", "text": prompt}]}]
|
168 |
-
prompt_full = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
169 |
-
inputs = processor(
|
170 |
-
text=[prompt_full],
|
171 |
-
images=[image],
|
172 |
-
return_tensors="pt",
|
173 |
-
padding=True,
|
174 |
-
truncation=True,
|
175 |
-
max_length=MAX_INPUT_TOKEN_LENGTH
|
176 |
-
).to(device)
|
177 |
-
|
178 |
-
# 3. Stream the generation
|
179 |
-
streamer = TextIteratorStreamer(processor, skip_prompt=True, skip_special_tokens=True)
|
180 |
-
generation_kwargs = {**inputs, "streamer": streamer, "max_new_tokens": max_new_tokens}
|
181 |
-
thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
182 |
-
thread.start()
|
183 |
-
|
184 |
-
buffer = ""
|
185 |
-
# Initial placeholder yield
|
186 |
-
yield buffer, "⏳ Generating response...", None
|
187 |
-
|
188 |
-
for new_text in streamer:
|
189 |
-
buffer += new_text
|
190 |
-
buffer = buffer.replace("<|im_end|>", "")
|
191 |
-
time.sleep(0.01) # Small delay for smoother streaming
|
192 |
-
yield buffer, "⏳ Generating response...", None
|
193 |
-
|
194 |
-
# 4. Process the final buffer content
|
195 |
try:
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
208 |
except Exception as e:
|
209 |
-
|
210 |
-
|
211 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
212 |
|
213 |
def create_gradio_interface():
|
214 |
-
"""Create the Gradio interface."""
|
215 |
css = """
|
216 |
.main-container { max-width: 1400px; margin: 0 auto; }
|
217 |
.header-text { text-align: center; color: #2c3e50; margin-bottom: 20px; }
|
218 |
.process-button {
|
219 |
-
border: none !important;
|
220 |
-
|
221 |
-
|
|
|
222 |
.process-button:hover {
|
223 |
-
background-color: darkblue !important;
|
224 |
-
|
225 |
-
|
|
|
|
|
|
|
|
|
226 |
"""
|
227 |
with gr.Blocks(theme="bethecloud/storj_theme", css=css) as demo:
|
228 |
gr.HTML("""
|
@@ -233,64 +361,72 @@ def create_gradio_interface():
|
|
233 |
</p>
|
234 |
</div>
|
235 |
""")
|
236 |
-
|
237 |
-
# Keep track of the uploaded image
|
238 |
-
image_state = gr.State(None)
|
239 |
-
|
240 |
with gr.Row():
|
241 |
-
# Left column - Input and controls
|
242 |
with gr.Column(scale=1):
|
243 |
model_choice = gr.Radio(
|
244 |
choices=["Camel-Doc-OCR-062825", "MonkeyOCR-Recognition", "Nanonets-OCR-s", "Megalodon-OCR-Sync-0713"],
|
245 |
label="Select Model",
|
246 |
value="Camel-Doc-OCR-062825"
|
247 |
)
|
248 |
-
file_input = gr.
|
249 |
label="Upload Image",
|
250 |
-
|
251 |
-
|
252 |
)
|
|
|
253 |
with gr.Accordion("Advanced Settings", open=False):
|
254 |
max_new_tokens = gr.Slider(minimum=1000, maximum=32000, value=24000, step=1000, label="Max New Tokens")
|
255 |
-
|
|
|
256 |
process_btn = gr.Button("🚀 Process Document", variant="primary", elem_classes=["process-button"], size="lg")
|
257 |
clear_btn = gr.Button("🗑️ Clear All", variant="secondary")
|
258 |
-
|
259 |
-
# Right column - Results
|
260 |
with gr.Column(scale=2):
|
261 |
with gr.Tabs():
|
262 |
with gr.Tab("📝 Extracted Content"):
|
263 |
-
|
264 |
-
with gr.Accordion("(
|
265 |
-
markdown_output = gr.Markdown(label="Formatted Result (Result.
|
266 |
-
|
267 |
with gr.Tab("📋 Layout JSON"):
|
268 |
-
json_output = gr.JSON(label="Layout Analysis Results
|
269 |
-
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
275 |
def clear_all():
|
276 |
-
"""
|
277 |
-
|
278 |
-
|
279 |
-
file_input.upload(handle_file_upload, inputs=[file_input], outputs=[image_state])
|
280 |
-
|
281 |
process_btn.click(
|
282 |
-
|
283 |
-
inputs=[model_choice,
|
284 |
-
outputs=[
|
285 |
)
|
286 |
-
|
287 |
clear_btn.click(
|
288 |
clear_all,
|
289 |
-
outputs=[file_input,
|
290 |
)
|
291 |
-
|
292 |
return demo
|
293 |
|
294 |
if __name__ == "__main__":
|
295 |
demo = create_gradio_interface()
|
296 |
-
demo.queue().launch(server_name="0.0.0.0", server_port=7860, share=True, show_error=True)
|
|
|
6 |
from io import BytesIO
|
7 |
from typing import Any, Dict, List, Optional, Tuple
|
8 |
import re
|
|
|
9 |
import time
|
10 |
+
from threading import Thread
|
11 |
|
12 |
import gradio as gr
|
13 |
import requests
|
14 |
import torch
|
15 |
+
from PIL import Image
|
16 |
from transformers import (
|
17 |
Qwen2_5_VLForConditionalGeneration,
|
18 |
AutoProcessor,
|
19 |
TextIteratorStreamer,
|
20 |
)
|
21 |
+
from qwen_vl_utils import process_vision_info
|
22 |
|
23 |
# Constants
|
24 |
MIN_PIXELS = 3136
|
25 |
MAX_PIXELS = 11289600
|
26 |
IMAGE_FACTOR = 28
|
27 |
+
MAX_INPUT_TOKEN_LENGTH = 2048
|
28 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
29 |
|
30 |
+
# Prompts
|
31 |
+
prompt = """Please output the layout information from the image, including each layout element's bbox, its category, and the corresponding text content within the bbox.
|
32 |
|
33 |
1. Bbox format: [x1, y1, x2, y2]
|
34 |
|
|
|
47 |
5. Final Output: The entire output must be a single JSON object.
|
48 |
"""
|
49 |
|
50 |
+
# Load models
|
51 |
MODEL_ID_M = "prithivMLmods/Camel-Doc-OCR-062825"
|
52 |
processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
|
53 |
model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
54 |
+
MODEL_ID_M,
|
55 |
+
trust_remote_code=True,
|
56 |
+
torch_dtype=torch.float16
|
57 |
).to(device).eval()
|
58 |
|
59 |
MODEL_ID_T = "prithivMLmods/Megalodon-OCR-Sync-0713"
|
60 |
processor_t = AutoProcessor.from_pretrained(MODEL_ID_T, trust_remote_code=True)
|
61 |
model_t = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
62 |
+
MODEL_ID_T,
|
63 |
+
trust_remote_code=True,
|
64 |
+
torch_dtype=torch.float16
|
65 |
).to(device).eval()
|
66 |
|
67 |
MODEL_ID_C = "nanonets/Nanonets-OCR-s"
|
68 |
processor_c = AutoProcessor.from_pretrained(MODEL_ID_C, trust_remote_code=True)
|
69 |
model_c = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
70 |
+
MODEL_ID_C,
|
71 |
+
trust_remote_code=True,
|
72 |
+
torch_dtype=torch.float16
|
73 |
).to(device).eval()
|
74 |
|
75 |
MODEL_ID_G = "echo840/MonkeyOCR"
|
76 |
SUBFOLDER = "Recognition"
|
77 |
processor_g = AutoProcessor.from_pretrained(
|
78 |
+
MODEL_ID_G,
|
79 |
+
trust_remote_code=True,
|
80 |
+
subfolder=SUBFOLDER
|
81 |
)
|
82 |
model_g = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
83 |
+
MODEL_ID_G,
|
84 |
+
trust_remote_code=True,
|
85 |
+
subfolder=SUBFOLDER,
|
86 |
+
torch_dtype=torch.float16
|
87 |
).to(device).eval()
|
88 |
|
|
|
89 |
# Utility functions
|
90 |
+
def round_by_factor(number: int, factor: int) -> int:
|
91 |
+
return round(number / factor) * factor
|
92 |
+
|
93 |
+
def smart_resize(
|
94 |
+
height: int,
|
95 |
+
width: int,
|
96 |
+
factor: int = 28,
|
97 |
+
min_pixels: int = 3136,
|
98 |
+
max_pixels: int = 11289600,
|
99 |
+
):
|
100 |
+
if max(height, width) / min(height, width) > 200:
|
101 |
+
raise ValueError(f"Aspect ratio too extreme: {max(height, width) / min(height, width)}")
|
102 |
+
h_bar = max(factor, round_by_factor(height, factor))
|
103 |
+
w_bar = max(factor, round_by_factor(width, factor))
|
104 |
+
if h_bar * w_bar > max_pixels:
|
105 |
+
beta = math.sqrt((height * width) / max_pixels)
|
106 |
+
h_bar = round_by_factor(height / beta, factor)
|
107 |
+
w_bar = round_by_factor(width / beta, factor)
|
108 |
+
elif h_bar * w_bar < min_pixels:
|
109 |
+
beta = math.sqrt(min_pixels / (height * width))
|
110 |
+
h_bar = round_by_factor(height * beta, factor)
|
111 |
+
w_bar = round_by_factor(width * beta, factor)
|
112 |
+
return h_bar, w_bar
|
113 |
+
|
114 |
+
def fetch_image(image_input, min_pixels: int = None, max_pixels: int = None):
|
115 |
+
if isinstance(image_input, str):
|
116 |
+
if image_input.startswith(("http://", "https://")):
|
117 |
+
response = requests.get(image_input)
|
118 |
+
image = Image.open(BytesIO(response.content)).convert('RGB')
|
119 |
+
else:
|
120 |
+
image = Image.open(image_input).convert('RGB')
|
121 |
+
elif isinstance(image_input, Image.Image):
|
122 |
+
image = image_input.convert('RGB')
|
123 |
+
else:
|
124 |
+
raise ValueError(f"Invalid image input type: {type(image_input)}")
|
125 |
+
if min_pixels or max_pixels:
|
126 |
+
min_pixels = min_pixels or MIN_PIXELS
|
127 |
+
max_pixels = max_pixels or MAX_PIXELS
|
128 |
+
height, width = smart_resize(image.height, image.width, factor=IMAGE_FACTOR, min_pixels=min_pixels, max_pixels=max_pixels)
|
129 |
+
image = image.resize((width, height), Image.LANCZOS)
|
130 |
+
return image
|
131 |
+
|
132 |
def is_arabic_text(text: str) -> bool:
|
|
|
133 |
if not text:
|
134 |
return False
|
135 |
+
header_pattern = r'^#{1,6}\s+(.+)$'
|
136 |
+
paragraph_pattern = r'^(?!#{1,6}\s|!\[|```|\||\s*[-*+]\s|\s*\d+\.\s)(.+)$'
|
137 |
+
content_text = []
|
138 |
+
for line in text.split('\n'):
|
139 |
+
line = line.strip()
|
140 |
+
if not line:
|
141 |
+
continue
|
142 |
+
header_match = re.match(header_pattern, line, re.MULTILINE)
|
143 |
+
if header_match:
|
144 |
+
content_text.append(header_match.group(1))
|
145 |
+
continue
|
146 |
+
if re.match(paragraph_pattern, line, re.MULTILINE):
|
147 |
+
content_text.append(line)
|
148 |
+
if not content_text:
|
149 |
+
return False
|
150 |
+
combined_text = ' '.join(content_text)
|
151 |
arabic_chars = 0
|
152 |
total_chars = 0
|
153 |
+
for char in combined_text:
|
154 |
if char.isalpha():
|
155 |
total_chars += 1
|
156 |
+
if ('\u0600' <= char <= '\u06FF') or ('\u0750' <= char <= '\u077F') or ('\u08A0' <= char <= '\u08FF'):
|
157 |
arabic_chars += 1
|
158 |
return total_chars > 0 and (arabic_chars / total_chars) > 0.5
|
159 |
|
160 |
def layoutjson2md(image: Image.Image, layout_data: List[Dict], text_key: str = 'text') -> str:
|
|
|
161 |
import base64
|
162 |
from io import BytesIO
|
163 |
markdown_lines = []
|
164 |
try:
|
|
|
165 |
sorted_items = sorted(layout_data, key=lambda x: (x.get('bbox', [0, 0, 0, 0])[1], x.get('bbox', [0, 0, 0, 0])[0]))
|
166 |
for item in sorted_items:
|
167 |
category = item.get('category', '')
|
168 |
text = item.get(text_key, '')
|
169 |
bbox = item.get('bbox', [])
|
|
|
170 |
if category == 'Picture':
|
171 |
if bbox and len(bbox) == 4:
|
172 |
try:
|
173 |
+
x1, y1, x2, y2 = bbox
|
174 |
+
x1, y1 = max(0, int(x1)), max(0, int(y1))
|
175 |
+
x2, y2 = min(image.width, int(x2)), min(image.height, int(y2))
|
176 |
+
if x2 > x1 and y2 > y1:
|
177 |
+
cropped_img = image.crop((x1, y1, x2, y2))
|
178 |
+
buffer = BytesIO()
|
179 |
+
cropped_img.save(buffer, format='PNG')
|
180 |
+
img_data = base64.b64encode(buffer.getvalue()).decode()
|
181 |
+
markdown_lines.append(f"<image-card alt="Image" src="data:image/png;base64,{img_data}" ></image-card>\n")
|
182 |
+
else:
|
183 |
+
markdown_lines.append("<image-card alt="Image" src="Image region detected" ></image-card>\n")
|
184 |
except Exception as e:
|
185 |
+
print(f"Error processing image region: {e}")
|
186 |
+
markdown_lines.append("<image-card alt="Image" src="Image detected" ></image-card>\n")
|
187 |
+
else:
|
188 |
+
markdown_lines.append("<image-card alt="Image" src="Image detected" ></image-card>\n")
|
189 |
elif not text:
|
190 |
continue
|
191 |
elif category == 'Title':
|
|
|
196 |
markdown_lines.append(f"{text}\n")
|
197 |
elif category == 'List-item':
|
198 |
markdown_lines.append(f"- {text}\n")
|
199 |
+
elif category == 'Table':
|
200 |
+
if text.strip().startswith('<'):
|
201 |
+
markdown_lines.append(f"{text}\n")
|
202 |
+
else:
|
203 |
+
markdown_lines.append(f"**Table:** {text}\n")
|
204 |
+
elif category == 'Formula':
|
205 |
+
if text.strip().startswith('$') or '\\' in text:
|
206 |
+
markdown_lines.append(f"$$ \n{text}\n $$\n")
|
207 |
+
else:
|
208 |
+
markdown_lines.append(f"**Formula:** {text}\n")
|
209 |
elif category == 'Caption':
|
210 |
markdown_lines.append(f"*{text}*\n")
|
211 |
elif category == 'Footnote':
|
212 |
+
markdown_lines.append(f"^{text}^\n")
|
213 |
+
elif category in ['Page-header', 'Page-footer']:
|
214 |
+
continue
|
215 |
+
else:
|
216 |
markdown_lines.append(f"{text}\n")
|
217 |
+
markdown_lines.append("")
|
218 |
except Exception as e:
|
219 |
print(f"Error converting to markdown: {e}")
|
220 |
+
return str(layout_data)
|
221 |
return "\n".join(markdown_lines)
|
222 |
|
|
|
223 |
@spaces.GPU
|
224 |
+
def inference(model_name: str, image: Image.Image, text: str, max_new_tokens: int = 1024) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
225 |
try:
|
226 |
+
if model_name == "Camel-Doc-OCR-062825":
|
227 |
+
processor = processor_m
|
228 |
+
model = model_m
|
229 |
+
elif model_name == "Megalodon-OCR-Sync-0713":
|
230 |
+
processor = processor_t
|
231 |
+
model = model_t
|
232 |
+
elif model_name == "Nanonets-OCR-s":
|
233 |
+
processor = processor_c
|
234 |
+
model = model_c
|
235 |
+
elif model_name == "MonkeyOCR-Recognition":
|
236 |
+
processor = processor_g
|
237 |
+
model = model_g
|
238 |
+
else:
|
239 |
+
raise ValueError(f"Invalid model selected: {model_name}")
|
240 |
+
|
241 |
+
if image is None:
|
242 |
+
yield "Please upload an image.", "Please upload an image."
|
243 |
+
return
|
244 |
+
|
245 |
+
messages = [{
|
246 |
+
"role": "user",
|
247 |
+
"content": [
|
248 |
+
{"type": "image", "image": image},
|
249 |
+
{"type": "text", "text": text},
|
250 |
+
]
|
251 |
+
}]
|
252 |
+
prompt_full = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
253 |
+
inputs = processor(
|
254 |
+
text=[prompt_full],
|
255 |
+
images=[image],
|
256 |
+
return_tensors="pt",
|
257 |
+
padding=True,
|
258 |
+
truncation=False,
|
259 |
+
max_length=MAX_INPUT_TOKEN_LENGTH
|
260 |
+
).to(device)
|
261 |
+
streamer = TextIteratorStreamer(processor, skip_prompt=True, skip_special_tokens=True)
|
262 |
+
generation_kwargs = {**inputs, "streamer": streamer, "max_new_tokens": max_new_tokens}
|
263 |
+
thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
264 |
+
thread.start()
|
265 |
+
buffer = ""
|
266 |
+
for new_text in streamer:
|
267 |
+
buffer += new_text
|
268 |
+
buffer = buffer.replace("<|im_end|>", "")
|
269 |
+
time.sleep(0.01)
|
270 |
+
yield buffer, buffer
|
271 |
except Exception as e:
|
272 |
+
print(f"Error during inference: {e}")
|
273 |
+
traceback.print_exc()
|
274 |
+
yield f"Error during inference: {str(e)}", f"Error during inference: {str(e)}"
|
275 |
+
|
276 |
+
def process_image(
|
277 |
+
model_name: str,
|
278 |
+
image: Image.Image,
|
279 |
+
min_pixels: Optional[int] = None,
|
280 |
+
max_pixels: Optional[int] = None,
|
281 |
+
max_new_tokens: int = 1024
|
282 |
+
) -> Dict[str, Any]:
|
283 |
+
try:
|
284 |
+
if min_pixels or max_pixels:
|
285 |
+
image = fetch_image(image, min_pixels=min_pixels, max_pixels=max_pixels)
|
286 |
+
result = {
|
287 |
+
'original_image': image,
|
288 |
+
'raw_output': "",
|
289 |
+
'layout_result': None,
|
290 |
+
'markdown_content': None
|
291 |
+
}
|
292 |
+
buffer = ""
|
293 |
+
for raw_output, _ in inference(model_name, image, prompt, max_new_tokens):
|
294 |
+
buffer = raw_output
|
295 |
+
result['raw_output'] = buffer
|
296 |
+
yield result
|
297 |
+
try:
|
298 |
+
json_match = re.search(r'```json
|
299 |
+
json_str = json_match.group(1) if json_match else buffer
|
300 |
+
layout_data = json.loads(json_str)
|
301 |
+
result['layout_result'] = layout_data
|
302 |
+
try:
|
303 |
+
markdown_content = layoutjson2md(image, layout_data, text_key='text')
|
304 |
+
result['markdown_content'] = markdown_content
|
305 |
+
except Exception as e:
|
306 |
+
print(f"Error generating markdown: {e}")
|
307 |
+
result['markdown_content'] = buffer
|
308 |
+
except json.JSONDecodeError:
|
309 |
+
print("Failed to parse JSON output, using raw output")
|
310 |
+
result['markdown_content'] = buffer
|
311 |
+
yield result
|
312 |
+
except Exception as e:
|
313 |
+
print(f"Error processing image: {e}")
|
314 |
+
traceback.print_exc()
|
315 |
+
result = {
|
316 |
+
'original_image': image,
|
317 |
+
'raw_output': f"Error processing image: {str(e)}",
|
318 |
+
'layout_result': None,
|
319 |
+
'markdown_content': f"Error processing image: {str(e)}"
|
320 |
+
}
|
321 |
+
yield result
|
322 |
+
|
323 |
+
def load_file_for_preview(file_path: str) -> Tuple[Optional[Image.Image], str]:
|
324 |
+
if not file_path or not os.path.exists(file_path):
|
325 |
+
return None, "No file selected"
|
326 |
+
file_ext = os.path.splitext(file_path)[1].lower()
|
327 |
+
try:
|
328 |
+
if file_ext in ['.jpg', '.jpeg', '.png', '.bmp', '.tiff']:
|
329 |
+
image = Image.open(file_path).convert('RGB')
|
330 |
+
return image, "Image loaded"
|
331 |
+
else:
|
332 |
+
return None, f"Unsupported file format: {file_ext}"
|
333 |
+
except Exception as e:
|
334 |
+
print(f"Error loading file: {e}")
|
335 |
+
return None, f"Error loading file: {str(e)}"
|
336 |
|
337 |
def create_gradio_interface():
|
|
|
338 |
css = """
|
339 |
.main-container { max-width: 1400px; margin: 0 auto; }
|
340 |
.header-text { text-align: center; color: #2c3e50; margin-bottom: 20px; }
|
341 |
.process-button {
|
342 |
+
border: none !important;
|
343 |
+
color: white !important;
|
344 |
+
font-weight: bold !important;
|
345 |
+
background-color: blue !important;}
|
346 |
.process-button:hover {
|
347 |
+
background-color: darkblue !important;
|
348 |
+
transform: translateY(-2px) !important;
|
349 |
+
box-shadow: 0 4px 8px rgba(0,0,0,0.2) !important; }
|
350 |
+
.info-box { border: 1px solid #dee2e6; border-radius: 8px; padding: 15px; margin: 10px 0; }
|
351 |
+
.page-info { text-align: center; padding: 8px 16px; border-radius: 20px; font-weight: bold; margin: 10px 0; }
|
352 |
+
.model-status { padding: 10px; border-radius: 8px; margin: 10px 0; text-align: center; font-weight: bold; }
|
353 |
+
.status-ready { background: #d1edff; color: #0c5460; border: 1px solid #b8daff; }
|
354 |
"""
|
355 |
with gr.Blocks(theme="bethecloud/storj_theme", css=css) as demo:
|
356 |
gr.HTML("""
|
|
|
361 |
</p>
|
362 |
</div>
|
363 |
""")
|
|
|
|
|
|
|
|
|
364 |
with gr.Row():
|
|
|
365 |
with gr.Column(scale=1):
|
366 |
model_choice = gr.Radio(
|
367 |
choices=["Camel-Doc-OCR-062825", "MonkeyOCR-Recognition", "Nanonets-OCR-s", "Megalodon-OCR-Sync-0713"],
|
368 |
label="Select Model",
|
369 |
value="Camel-Doc-OCR-062825"
|
370 |
)
|
371 |
+
file_input = gr.File(
|
372 |
label="Upload Image",
|
373 |
+
file_types=[".jpg", ".jpeg", ".png", ".bmp", ".tiff"],
|
374 |
+
type="filepath"
|
375 |
)
|
376 |
+
image_preview = gr.Image(label="Preview", type="pil", interactive=False, height=300)
|
377 |
with gr.Accordion("Advanced Settings", open=False):
|
378 |
max_new_tokens = gr.Slider(minimum=1000, maximum=32000, value=24000, step=1000, label="Max New Tokens")
|
379 |
+
min_pixels = gr.Number(value=MIN_PIXELS, label="Min Pixels")
|
380 |
+
max_pixels = gr.Number(value=MAX_PIXELS, label="Max Pixels")
|
381 |
process_btn = gr.Button("🚀 Process Document", variant="primary", elem_classes=["process-button"], size="lg")
|
382 |
clear_btn = gr.Button("🗑️ Clear All", variant="secondary")
|
|
|
|
|
383 |
with gr.Column(scale=2):
|
384 |
with gr.Tabs():
|
385 |
with gr.Tab("📝 Extracted Content"):
|
386 |
+
output = gr.Textbox(label="Raw Output Stream", interactive=False, lines=2, show_copy_button=True)
|
387 |
+
with gr.Accordion("(Result.md)", open=False):
|
388 |
+
markdown_output = gr.Markdown(label="Formatted Result (Result.Md)")
|
|
|
389 |
with gr.Tab("📋 Layout JSON"):
|
390 |
+
json_output = gr.JSON(label="Layout Analysis Results", value=None)
|
391 |
+
def process_document(model_name, file_path, max_tokens, min_pix, max_pix):
|
392 |
+
try:
|
393 |
+
if not file_path:
|
394 |
+
return "Please upload an image.", "Please upload an image.", None
|
395 |
+
image, status = load_file_for_preview(file_path)
|
396 |
+
if image is None:
|
397 |
+
return status, status, None
|
398 |
+
for result in process_image(model_name, image, min_pixels=int(min_pix) if min_pix else None, max_pixels=int(max_pix) if max_pix else None, max_new_tokens=max_tokens):
|
399 |
+
raw_output = result['raw_output']
|
400 |
+
markdown_content = result['markdown_content'] or raw_output
|
401 |
+
if is_arabic_text(markdown_content):
|
402 |
+
markdown_update = gr.update(value=markdown_content, rtl=True)
|
403 |
+
else:
|
404 |
+
markdown_update = markdown_content
|
405 |
+
yield raw_output, markdown_update, result['layout_result']
|
406 |
+
except Exception as e:
|
407 |
+
error_msg = f"Error processing document: {str(e)}"
|
408 |
+
print(error_msg)
|
409 |
+
traceback.print_exc()
|
410 |
+
yield error_msg, error_msg, None
|
411 |
+
def handle_file_upload(file_path):
|
412 |
+
if not file_path:
|
413 |
+
return None, "No file loaded"
|
414 |
+
image, page_info = load_file_for_preview(file_path)
|
415 |
+
return image, page_info
|
416 |
def clear_all():
|
417 |
+
return None, None, "No file loaded", "", "Click 'Process Document' to see extracted content...", None
|
418 |
+
file_input.change(handle_file_upload, inputs=[file_input], outputs=[image_preview, output])
|
|
|
|
|
|
|
419 |
process_btn.click(
|
420 |
+
process_document,
|
421 |
+
inputs=[model_choice, file_input, max_new_tokens, min_pixels, max_pixels],
|
422 |
+
outputs=[output, markdown_output, json_output]
|
423 |
)
|
|
|
424 |
clear_btn.click(
|
425 |
clear_all,
|
426 |
+
outputs=[file_input, image_preview, output, markdown_output, json_output]
|
427 |
)
|
|
|
428 |
return demo
|
429 |
|
430 |
if __name__ == "__main__":
|
431 |
demo = create_gradio_interface()
|
432 |
+
demo.queue(max_size=10).launch(server_name="0.0.0.0", server_port=7860, share=False, debug=True, show_error=True)
|