Update app.py

app.py

@@ -35,7 +35,7 @@ def create_monitor_interface():
         def analyze_frame(self, frame: np.ndarray) -> str:
             if frame is None:
                 return "No frame received"
-                
+                    
             # Convert and resize image
             if len(frame.shape) == 2:
                 frame = cv2.cvtColor(frame, cv2.COLOR_GRAY2RGB)
@@ -45,12 +45,12 @@ def create_monitor_interface():
             frame = self.resize_image(frame)
             frame_pil = PILImage.fromarray(frame)
             
-            # Convert to base64 with minimal quality
+            # High quality image for better analysis
             buffered = io.BytesIO()
             frame_pil.save(buffered, 
-                         format="JPEG", 
-                         quality=30, 
-                         optimize=True)
+                          format="JPEG", 
+                          quality=95,
+                          optimize=True)
             img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
             image_url = f"data:image/jpeg;base64,{img_base64}"
             
@@ -58,14 +58,43 @@ def create_monitor_interface():
                 completion = self.client.chat.completions.create(
                     model=self.model_name,
                     messages=[
+                        {
+                            "role": "system",
+                            "content": """You are a comprehensive safety analysis system. Analyze images for ALL types of safety concerns including but not limited to:
+                            - Personal Protective Equipment (PPE)
+                            - Ergonomic issues
+                            - Fire and electrical hazards
+                            - Chemical and environmental hazards
+                            - Machine and equipment safety
+                            - Fall protection and working at heights
+                            - Material handling and storage
+                            - Emergency access and exits
+                            - Housekeeping and organization
+                            - Lighting and visibility
+                            - Ventilation and air quality
+                            - Tool safety and maintenance"""
+                        },
                         {
                             "role": "user",
                             "content": [
                                 {
                                     "type": "text",
-                                    "text": """Analyze this workplace image and describe each safety concern in this format:
-                                    - <location>Description</location>
-                                    Use one line per issue, starting with a dash and location in tags."""
+                                    "text": """Analyze this image for ANY safety concerns or hazards. For each issue identified, specify:
+        
+        1. The exact location in the image (be specific: top-left, center-right, bottom, etc.)
+        2. The type of safety concern
+        3. The potential risk or hazard
+        4. Any relevant safety standards being violated
+        
+        Format each observation as:
+        - <location>position:safety issue description</location>
+        
+        Example formats:
+        - <location>top-right:Exposed electrical wiring creating shock hazard</location>
+        - <location>bottom-left:Improperly stored chemicals without proper labeling</location>
+        - <location>center:Missing machine guarding on rotating equipment</location>
+        
+        Be thorough and identify ALL safety issues, not just the obvious ones."""
                                 },
                                 {
                                     "type": "image_url",
@@ -74,17 +103,11 @@ def create_monitor_interface():
                                     }
                                 }
                             ]
-                        },
-                        {
-                            "role": "assistant",
-                            "content": ""
                         }
                     ],
-                    temperature=0.1,
-                    max_tokens=150,
-                    top_p=1,
-                    stream=False,
-                    stop=None
+                    temperature=0.7,  # Higher temperature for more comprehensive analysis
+                    max_tokens=500,
+                    stream=False
                 )
                 return completion.choices[0].message.content
             except Exception as e:
@@ -92,29 +115,77 @@ def create_monitor_interface():
                 return f"Analysis Error: {str(e)}"
 
         def draw_observations(self, image, observations):
+            """Draw accurate bounding boxes based on safety issue locations."""
             height, width = image.shape[:2]
             font = cv2.FONT_HERSHEY_SIMPLEX
             font_scale = 0.5
             thickness = 2
+            padding = 10
             
-            # Generate random positions for each observation
-            for idx, obs in enumerate(observations):
-                color = self.colors[idx % len(self.colors)]
+            def get_region_coordinates(position: str) -> tuple:
+                """Get coordinates based on position description."""
+                # Basic regions
+                regions = {
+                    'top-left': (0, 0, width//3, height//3),
+                    'top': (width//3, 0, 2*width//3, height//3),
+                    'top-right': (2*width//3, 0, width, height//3),
+                    'center-left': (0, height//3, width//3, 2*height//3),
+                    'center': (width//3, height//3, 2*width//3, 2*height//3),
+                    'center-right': (2*width//3, height//3, width, 2*height//3),
+                    'bottom-left': (0, 2*height//3, width//3, height),
+                    'bottom': (width//3, 2*height//3, 2*width//3, height),
+                    'bottom-right': (2*width//3, 2*height//3, width, height),
+                    'left': (0, height//4, width//3, 3*height//4),
+                    'right': (2*width//3, height//4, width, 3*height//4)
+                }
+                
+                # Find best matching region
+                best_match = 'center'
+                max_words = 0
+                pos_lower = position.lower()
                 
-                # Generate random box position
-                box_width = width // 3
-                box_height = height // 3
-                x = random.randint(0, width - box_width)
-                y = random.randint(0, height - box_height)
+                for region in regions.keys():
+                    words = region.split('-')
+                    matches = sum(1 for word in words if word in pos_lower)
+                    if matches > max_words:
+                        max_words = matches
+                        best_match = region
                 
-                # Draw rectangle
-                cv2.rectangle(image, (x, y), (x + box_width, y + box_height), color, 2)
+                return regions[best_match]
+            
+            for idx, obs in enumerate(observations):
+                color = self.colors[idx % len(self.colors)]
                 
-                # Add label with background
-                label = obs[:40] + "..." if len(obs) > 40 else obs
-                label_size = cv2.getTextSize(label, font, font_scale, thickness)[0]
-                cv2.rectangle(image, (x, y - 20), (x + label_size[0], y), color, -1)
-                cv2.putText(image, label, (x, y - 5), font, font_scale, (255, 255, 255), thickness)
+                # Parse location and description
+                parts = obs.split(':')
+                if len(parts) >= 2:
+                    position = parts[0]
+                    description = ':'.join(parts[1:])
+                    
+                    # Get region coordinates
+                    x1, y1, x2, y2 = get_region_coordinates(position)
+                    
+                    # 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)
+                    
+                    # Position text above the box
+                    text_x = max(0, x1)
+                    text_y = max(label_size[1] + padding, y1 - padding)
+                    
+                    # Draw text background
+                    cv2.rectangle(image, 
+                                 (text_x, text_y - label_size[1] - padding),
+                                 (text_x + label_size[0] + padding, text_y),
+                                 color, -1)
+                    
+                    # Draw text
+                    cv2.putText(image, label,
+                                (text_x + padding//2, text_y - padding//2),
+                                font, font_scale, (255, 255, 255), thickness)
             
             return image