Update app.py

app.py

@@ -18,20 +18,6 @@ def create_monitor_interface():
             self.max_image_size = (800, 800)
             self.colors = [(0, 0, 255), (255, 0, 0), (0, 255, 0), (255, 255, 0), (255, 0, 255)]
             
-        def resize_image(self, image):
-            height, width = image.shape[:2]
-            
-            if height > self.max_image_size[1] or width > self.max_image_size[0]:
-                aspect = width / height
-                if width > height:
-                    new_width = self.max_image_size[0]
-                    new_height = int(new_width / aspect)
-                else:
-                    new_height = self.max_image_size[1]
-                    new_width = int(new_height * aspect)
-                return cv2.resize(image, (new_width, new_height), interpolation=cv2.INTER_AREA)
-            return image
-
         def analyze_frame(self, frame: np.ndarray) -> str:
             if frame is None:
                 return ""
@@ -57,19 +43,29 @@ def create_monitor_interface():
                 completion = self.client.chat.completions.create(
                     model=self.model_name,
                     messages=[
+                        {
+                            "role": "system",
+                            "content": "You are a safety analysis expert. Analyze images for safety concerns and provide detailed observations."
+                        },
                         {
                             "role": "user",
                             "content": [
                                 {
                                     "type": "text",
-                                    "text": """Analyze this image for safety concerns. For each specific issue you identify, provide:
-                                    1. Exact location in the image (e.g., 'top-left', 'center', 'bottom-right', etc.)
-                                    2. Description of the safety concern
-                                    
-                                    Format your response with each issue on a new line as:
-                                    - <location>position:detailed description of the safety concern</location>
-                                    
-                                    Be specific about what you observe in the image."""
+                                    "text": """Analyze this image for safety concerns and risks. For each issue you identify:
+
+1. Specify the exact location in the image where the issue is visible
+2. Describe what the safety concern is
+3. Include any relevant details about PPE, posture, equipment, or environmental hazards
+
+Format EACH observation exactly like this:
+- <location>position:detailed description of the concern</location>
+
+Example format:
+- <location>center:Worker bending incorrectly while lifting heavy materials</location>
+- <location>top-right:Missing safety guardrail near elevated platform</location>
+
+Provide multiple observations if you see multiple issues."""
                                 },
                                 {
                                     "type": "image_url",
@@ -78,23 +74,34 @@ def create_monitor_interface():
                                     }
                                 }
                             ]
-                        },
-                        {
-                            "role": "assistant",
-                            "content": ""
                         }
                     ],
-                    temperature=0.2,
+                    temperature=0.5,  # Increased for more varied observations
                     max_tokens=500,
-                    top_p=1,
-                    stream=False,
-                    stop=None
+                    stream=False
                 )
-                return completion.choices[0].message.content
+                
+                response = completion.choices[0].message.content
+                print(f"Raw response: {response}")  # For debugging
+                return response
+                
             except Exception as e:
                 print(f"Analysis error: {str(e)}")
                 return ""
 
+        def resize_image(self, image):
+            height, width = image.shape[:2]
+            if height > self.max_image_size[1] or width > self.max_image_size[0]:
+                aspect = width / height
+                if width > height:
+                    new_width = self.max_image_size[0]
+                    new_height = int(new_width / aspect)
+                else:
+                    new_height = self.max_image_size[1]
+                    new_width = int(new_height * aspect)
+                return cv2.resize(image, (new_width, new_height), interpolation=cv2.INTER_AREA)
+            return image
+
         def get_region_coordinates(self, position: str, image_shape: tuple) -> tuple:
             height, width = image_shape[:2]
             regions = {
@@ -109,10 +116,19 @@ def create_monitor_interface():
                 'bottom-right': (2*width//3, 2*height//3, width, height)
             }
             
-            for region_name, coords in regions.items():
-                if region_name in position.lower():
-                    return coords
+            # Try to match the position with regions
+            matched_region = None
+            max_match_length = 0
+            position_lower = position.lower()
+            
+            for region_name in regions:
+                if region_name in position_lower:
+                    if len(region_name) > max_match_length:
+                        matched_region = region_name
+                        max_match_length = len(region_name)
             
+            if matched_region:
+                return regions[matched_region]
             return regions['center']
 
         def draw_observations(self, image, observations):
@@ -128,27 +144,23 @@ def create_monitor_interface():
                 if len(parts) >= 2:
                     position = parts[0]
                     description = ':'.join(parts[1:])
-                else:
-                    continue
-                
-                x1, y1, x2, y2 = self.get_region_coordinates(position, image.shape)
-                
-                # Draw rectangle
-                cv2.rectangle(image, (x1, y1), (x2, y2), color, 2)
-                
-                # Add label with background
-                label = description[:50] + "..." if len(description) > 50 else description
-                label_size = cv2.getTextSize(label, font, font_scale, thickness)[0]
-                
-                label_x = max(0, min(x1, width - label_size[0]))
-                label_y = max(20, y1 - 5)
-                
-                # Draw background for text
-                cv2.rectangle(image, (label_x, label_y - 20), 
-                            (label_x + label_size[0], label_y), color, -1)
-                # Draw text
-                cv2.putText(image, label, (label_x, label_y - 5), 
-                          font, font_scale, (255, 255, 255), thickness)
+                    
+                    x1, y1, x2, y2 = self.get_region_coordinates(position, image.shape)
+                    
+                    # Draw rectangle
+                    cv2.rectangle(image, (x1, y1), (x2, y2), color, 2)
+                    
+                    # Add label with background
+                    label = description[:50] + "..." if len(description) > 50 else description
+                    label_size = cv2.getTextSize(label, font, font_scale, thickness)[0]
+                    
+                    label_x = max(0, min(x1, width - label_size[0]))
+                    label_y = max(20, y1 - 5)
+                    
+                    cv2.rectangle(image, (label_x, label_y - 20), 
+                                (label_x + label_size[0], label_y), color, -1)
+                    cv2.putText(image, label, (label_x, label_y - 5), 
+                              font, font_scale, (255, 255, 255), thickness)
             
             return image
 
@@ -157,8 +169,8 @@ def create_monitor_interface():
                 return None, "No image provided"
             
             analysis = self.analyze_frame(frame)
+            print(f"Analysis received: {analysis}")  # Debug print
             
-            # Parse observations
             observations = []
             for line in analysis.split('\n'):
                 line = line.strip()
@@ -170,14 +182,19 @@ def create_monitor_interface():
                         if observation and ':' in observation:
                             observations.append(observation)
             
+            print(f"Parsed observations: {observations}")  # Debug print
+            
             display_frame = frame.copy()
             if observations:
                 annotated_frame = self.draw_observations(display_frame, observations)
                 return annotated_frame, analysis
-            else:
-                return display_frame, "No safety concerns detected in the image."
+            
+            # If no observations were found but we got some analysis
+            if analysis and not analysis.isspace():
+                return display_frame, analysis
+            
+            return display_frame, "Please try again - no safety analysis was generated."
 
-    # Create the main interface
     monitor = SafetyMonitor()
     
     with gr.Blocks() as demo:
@@ -205,13 +222,6 @@ def create_monitor_interface():
             outputs=[output_image, analysis_text]
         )
 
-        gr.Markdown("""
-        ## Instructions:
-        1. Upload an image to analyze
-        2. View identified safety concerns with bounding boxes
-        3. Read detailed analysis results
-        """)
-
     return demo
 
 demo = create_monitor_interface()