Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -58,35 +58,55 @@ def extract_metadata(image):
|
|
| 58 |
|
| 59 |
try:
|
| 60 |
# Ensure image is PIL Image
|
|
|
|
| 61 |
if not isinstance(image, PILImage.Image):
|
| 62 |
image = PILImage.open(image).convert("RGB")
|
|
|
|
|
|
|
| 63 |
|
| 64 |
# Set format (required by Outlines Image class)
|
| 65 |
if not image.format:
|
| 66 |
image.format = "PNG"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
|
| 68 |
# Create Chat prompt with Image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
prompt = Chat(
|
| 70 |
messages=[
|
| 71 |
{
|
| 72 |
"role": "user",
|
| 73 |
-
"content":
|
| 74 |
-
{"type": "text", "text": EXTRACTION_PROMPT},
|
| 75 |
-
{"type": "image", "image": Image(image)},
|
| 76 |
-
],
|
| 77 |
}
|
| 78 |
]
|
| 79 |
)
|
|
|
|
| 80 |
|
| 81 |
# Generate with structured output - guaranteed valid JSON
|
|
|
|
| 82 |
result = model(prompt, CatalogCardMetadata, max_new_tokens=512)
|
|
|
|
| 83 |
|
| 84 |
# Parse and format (always valid JSON with Outlines)
|
| 85 |
metadata = CatalogCardMetadata.model_validate_json(result)
|
| 86 |
-
return
|
| 87 |
|
| 88 |
except Exception as e:
|
| 89 |
-
|
|
|
|
|
|
|
|
|
|
| 90 |
|
| 91 |
|
| 92 |
# Create Gradio interface
|
|
@@ -111,7 +131,7 @@ with gr.Blocks(title="Library Card Metadata Extractor") as demo:
|
|
| 111 |
|
| 112 |
with gr.Column(scale=1):
|
| 113 |
gr.Markdown("### π Extracted Metadata (JSON)")
|
| 114 |
-
output = gr.
|
| 115 |
|
| 116 |
submit_btn.click(fn=extract_metadata, inputs=image_input, outputs=output)
|
| 117 |
|
|
|
|
| 58 |
|
| 59 |
try:
|
| 60 |
# Ensure image is PIL Image
|
| 61 |
+
print(f"DEBUG: Received image type: {type(image)}")
|
| 62 |
if not isinstance(image, PILImage.Image):
|
| 63 |
image = PILImage.open(image).convert("RGB")
|
| 64 |
+
print(f"DEBUG: After conversion, image type: {type(image)}")
|
| 65 |
+
print(f"DEBUG: Image format before setting: {image.format}")
|
| 66 |
|
| 67 |
# Set format (required by Outlines Image class)
|
| 68 |
if not image.format:
|
| 69 |
image.format = "PNG"
|
| 70 |
+
print(f"DEBUG: Image format after setting: {image.format}")
|
| 71 |
+
|
| 72 |
+
# Wrap in Outlines Image
|
| 73 |
+
outlines_image = Image(image)
|
| 74 |
+
print(f"DEBUG: Outlines Image created: {type(outlines_image)}")
|
| 75 |
+
print(f"DEBUG: Outlines Image.image type: {type(outlines_image.image)}")
|
| 76 |
|
| 77 |
# Create Chat prompt with Image
|
| 78 |
+
content_list = [
|
| 79 |
+
{"type": "text", "text": EXTRACTION_PROMPT},
|
| 80 |
+
{"type": "image", "image": outlines_image},
|
| 81 |
+
]
|
| 82 |
+
print(f"DEBUG: Content list created: {len(content_list)} items")
|
| 83 |
+
print(f"DEBUG: Content[0] type: {content_list[0]['type']}")
|
| 84 |
+
print(f"DEBUG: Content[1] type: {content_list[1]['type']}, image type: {type(content_list[1]['image'])}")
|
| 85 |
+
|
| 86 |
prompt = Chat(
|
| 87 |
messages=[
|
| 88 |
{
|
| 89 |
"role": "user",
|
| 90 |
+
"content": content_list,
|
|
|
|
|
|
|
|
|
|
| 91 |
}
|
| 92 |
]
|
| 93 |
)
|
| 94 |
+
print(f"DEBUG: Chat prompt created successfully")
|
| 95 |
|
| 96 |
# Generate with structured output - guaranteed valid JSON
|
| 97 |
+
print(f"DEBUG: Starting generation...")
|
| 98 |
result = model(prompt, CatalogCardMetadata, max_new_tokens=512)
|
| 99 |
+
print(f"DEBUG: Generation complete, result type: {type(result)}")
|
| 100 |
|
| 101 |
# Parse and format (always valid JSON with Outlines)
|
| 102 |
metadata = CatalogCardMetadata.model_validate_json(result)
|
| 103 |
+
return json.dumps(metadata.model_dump(exclude_none=True), indent=2)
|
| 104 |
|
| 105 |
except Exception as e:
|
| 106 |
+
import traceback
|
| 107 |
+
error_msg = f"Error during extraction: {str(e)}\n\nTraceback:\n{traceback.format_exc()}"
|
| 108 |
+
print(error_msg)
|
| 109 |
+
return error_msg
|
| 110 |
|
| 111 |
|
| 112 |
# Create Gradio interface
|
|
|
|
| 131 |
|
| 132 |
with gr.Column(scale=1):
|
| 133 |
gr.Markdown("### π Extracted Metadata (JSON)")
|
| 134 |
+
output = gr.Code(label="Metadata", language="json", lines=15)
|
| 135 |
|
| 136 |
submit_btn.click(fn=extract_metadata, inputs=image_input, outputs=output)
|
| 137 |
|