Create app.py

app.py

@@ -0,0 +1,128 @@
+# Copyright (c) Alibaba Cloud.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+import os
+import re
+import gradio as gr
+import dashscope
+from dashscope import MultiModalConversation
+from argparse import ArgumentParser
+from http import HTTPStatus
+from urllib3.exceptions import HTTPError
+
+# Set API key
+API_KEY = os.environ['API_KEY']
+dashscope.api_key = API_KEY
+
+# Constants
+REVISION = 'v1.0.4'
+BOX_TAG_PATTERN = r"<box>\((\d+),(\d+),(\d+),(\d+)\)</box>"
+
+def _get_args():
+    parser = ArgumentParser()
+    parser.add_argument("--revision", type=str, default=REVISION)
+    parser.add_argument("--share", action="store_true", default=False,
+                        help="Create a publicly shareable link for the interface.")
+    parser.add_argument("--server-port", type=int, default=7860,
+                        help="Demo server port.")
+    parser.add_argument("--server-name", type=str, default="127.0.0.1",
+                        help="Demo server name.")
+    return parser.parse_args()
+
+def parse_bounding_boxes(text):
+    """Parse bounding box coordinates from model output."""
+    matches = re.findall(BOX_TAG_PATTERN, text)
+    bboxes = []
+    for match in matches:
+        x1, y1, x2, y2 = map(int, match)
+        bboxes.append(((x1, y1, x2, y2), "Object"))
+    return bboxes
+
+def predict(image, prompt):
+    """Process image and prompt to get bounding boxes and return annotated image."""
+    if image is None or not prompt:
+        return None, "Please upload an image and provide a prompt."
+
+    # Prepare message for the model
+    messages = [{
+        'role': 'user',
+        'content': [
+            {'image': f'file://{image}'},
+            {'text': prompt}
+        ]
+    }]
+
+    # Call the Qwen2.5-VL model
+    try:
+        responses = MultiModalConversation.call(
+            model='qwen2.5-vl-32b-instruct',
+            messages=messages,
+            stream=False
+        )
+        if responses.status_code != HTTPStatus.OK:
+            return None, f"Error: {responses.message}"
+
+        # Extract response text
+        response = responses.output.choices[0].message.content
+        response_text = ''.join([ele['text'] if 'text' in ele else ele.get('box', '') for ele in response])
+
+        # Parse bounding boxes
+        bboxes = parse_bounding_boxes(response_text)
+        if not bboxes:
+            return None, "No bounding boxes detected."
+
+        # Return the image and annotations for AnnotatedImage
+        return (image, bboxes), None
+
+    except HTTPError as e:
+        return None, f"HTTP Error: {str(e)}"
+    except Exception as e:
+        return None, f"Error: {str(e)}"
+
+def clear_inputs():
+    """Reset the input fields."""
+    return None, "", None
+
+def _launch_demo(args):
+    with gr.Blocks() as demo:
+        gr.Markdown("""<center><font size=3> Qwen2.5-VL-32B-Instruct Bounding Box Demo </center>""")
+
+        with gr.Row():
+            with gr.Column():
+                image_input = gr.Image(type="filepath", label="Upload Image")
+                prompt_input = gr.Textbox(
+                    lines=2,
+                    label="Prompt",
+                    value="Detect all objects in the provided image and output their bounding box coordinates in the format <box>(x1,y1,x2,y2)</box>. Do not include any other text or descriptions. If multiple objects are detected, list each bounding box in a new <box> tag."
+                )
+                submit_btn = gr.Button("🚀 Submit")
+                clear_btn = gr.Button("🧹 Clear")
+            with gr.Column():
+                output_image = gr.AnnotatedImage(label="Annotated Image")
+                error_message = gr.Textbox(label="Status", interactive=False)
+
+        # Bind actions
+        submit_btn.click(
+            fn=predict,
+            inputs=[image_input, prompt_input],
+            outputs=[output_image, error_message]
+        )
+        clear_btn.click(
+            fn=clear_inputs,
+            inputs=[],
+            outputs=[image_input, prompt_input, output_image, error_message]
+        )
+
+    demo.launch(
+        share=args.share,
+        server_port=args.server_port,
+        server_name=args.server_name
+    )
+
+def main():
+    args = _get_args()
+    _launch_demo(args)
+
+if __name__ == '__main__':
+    main()