Update app.py

app.py

@@ -27,7 +27,6 @@ def analyze_image(image, prompt):
             {"type": "text", "text": prompt}
         ]}
     ]
-    
     input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
     inputs = processor(
         image,
@@ -35,41 +34,39 @@ def analyze_image(image, prompt):
         add_special_tokens=False,
         return_tensors="pt"
     ).to(model.device)
-
+    
     with torch.no_grad():
         output = model.generate(**inputs, max_new_tokens=100)
     
     full_response = processor.decode(output[0])
-    print("Full response:", full_response)  # Debug print
-
-    # return full_response
+    
     try:
-        json_match = re.search(r'\{.*?\}', full_response, re.DOTALL)
-        if json_match:
-            json_str = json_match.group(0)
+        # Find all JSON-like structures in the response
+        json_matches = list(re.finditer(r'\{.*?\}', full_response, re.DOTALL))
+        
+        if json_matches:
+            # Take the last match
+            last_json_str = json_matches[-1].group(0)
             try:
-                return json.loads(json_str)
+                processed_json = json.loads(last_json_str)
             except json.JSONDecodeError as e:
-                print(f"JSON decode error: {e}")
-                return {"error": "Invalid JSON in model output", "full_response": full_response}
+                processed_json = {"error": f"Invalid JSON in model output: {e}", "full_response": full_response}
         else:
-            return {"error": "No JSON found in model output", "full_response": full_response}
+            processed_json = {"error": "No JSON found in model output", "full_response": full_response}
     except Exception as e:
-        print(f"Error in analyze_image: {e}")
-        return {"Full Response": str(e), "full_response": full_response}
-
+        processed_json = {"error": str(e), "full_response": full_response}
+    
+    return full_response, processed_json
 
-default_prompt = """Analyze this image and determine if it contains a data logger. 
-A data logger is typically a small, black electronic device used to monitor and record data 
-over time, such as voltage, temperature, or current, via external sensors.
+default_prompt = """Analyze this image and determine if it contains a data logger. A data logger is typically a small, black electronic device used to monitor and record data over time, such as voltage, temperature, or current, via external sensors.
 
-If a data logger is present in the image, respond with:
-{"present": true, "reason": "Brief explanation of why you believe it's a data logger"}
+Carefully examine the image and provide a detailed response. If a data logger is present in the image, respond with:
+{"present": true, "reason": "Detailed explanation of why you believe it's a data logger, including specific visual cues you've identified"}
 
 If no data logger is visible, respond with:
-{"present": false, "reason": "Brief explanation of why you believe there's no data logger"}
+{"present": false, "reason": "Detailed explanation of why you believe there's no data logger, describing what you see instead"}
 
-Ensure your response is in valid JSON format."""
+Be specific in your reasoning, mentioning colors, shapes, and other relevant details you observe in the image. Ensure your response is in valid JSON format and provides substantial, meaningful analysis."""
 
 iface = gr.Interface(
     fn=analyze_image,
@@ -77,12 +74,104 @@ iface = gr.Interface(
         gr.Image(type="pil", label="Upload Image"),
         gr.Textbox(label="Prompt", value=default_prompt, lines=10)
     ],
-    outputs=gr.JSON(label="Analysis Result"),
-    title="Data Logger Detection using Llama 3.2 Vision",
-    description="Upload an image and customize the prompt to check if it contains a data logger.",
+    outputs=[
+        gr.Textbox(label="Full Response", lines=10),
+        gr.JSON(label="Processed JSON")
+    ],
+    title="Llama 3.2 Vision",
+    description=" ",
     examples=[
-        ["bad.png", default_prompt]
+        ["./bad.jpg", default_prompt]  
     ]
 )
 
-iface.launch()
+iface.launch()
+
+# import torch
+# from PIL import Image
+# from transformers import AutoProcessor, AutoModelForPreTraining
+# import gradio as gr
+# import json
+# import traceback
+# import os
+# import re
+
+# model_name = "meta-llama/Llama-3.2-11B-Vision-Instruct"
+# token = os.getenv("HUGGINGFACE_TOKEN").strip()
+
+# processor = AutoProcessor.from_pretrained(model_name, token=token)
+# model = AutoModelForPreTraining.from_pretrained(
+#     model_name,
+#     quantization_config={"load_in_4bit": True},
+#     token=token
+# )
+
+# if torch.cuda.is_available():
+#     model = model.to('cuda')
+
+# def analyze_image(image, prompt):
+#     messages = [
+#         {"role": "user", "content": [
+#             {"type": "image"},
+#             {"type": "text", "text": prompt}
+#         ]}
+#     ]
+    
+#     input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
+#     inputs = processor(
+#         image,
+#         input_text,
+#         add_special_tokens=False,
+#         return_tensors="pt"
+#     ).to(model.device)
+
+#     with torch.no_grad():
+#         output = model.generate(**inputs, max_new_tokens=100)
+    
+#     full_response = processor.decode(output[0])
+#     print("Full response:", full_response)  # Debug print
+
+#     # return full_response
+#     try:
+#         json_match = re.search(r'\{.*?\}', full_response, re.DOTALL)
+#         if json_match:
+#             json_str = json_match.group(0)
+#             try:
+#                 return json.loads(json_str)
+#             except json.JSONDecodeError as e:
+#                 print(f"JSON decode error: {e}")
+#                 return {"error": "Invalid JSON in model output", "full_response": full_response}
+#         else:
+#             return {"error": "No JSON found in model output", "full_response": full_response}
+#     except Exception as e:
+#         print(f"Error in analyze_image: {e}")
+#         return {"Full Response": str(e), "full_response": full_response}
+
+
+# default_prompt = """Analyze this image and determine if it contains a data logger. 
+# A data logger is typically a small, black electronic device used to monitor and record data 
+# over time, such as voltage, temperature, or current, via external sensors.
+
+# If a data logger is present in the image, respond with:
+# {"present": true, "reason": "Brief explanation of why you believe it's a data logger"}
+
+# If no data logger is visible, respond with:
+# {"present": false, "reason": "Brief explanation of why you believe there's no data logger"}
+
+# Ensure your response is in valid JSON format."""
+
+# iface = gr.Interface(
+#     fn=analyze_image,
+#     inputs=[
+#         gr.Image(type="pil", label="Upload Image"),
+#         gr.Textbox(label="Prompt", value=default_prompt, lines=10)
+#     ],
+#     outputs=gr.JSON(label="Analysis Result"),
+#     title="Data Logger Detection using Llama 3.2 Vision",
+#     description="Upload an image and customize the prompt to check if it contains a data logger.",
+#     examples=[
+#         ["bad.png", default_prompt]
+#     ]
+# )
+
+# iface.launch()