Update app.py

app.py

@@ -289,7 +289,7 @@ def analyze_outfit(input_img):
             # Continue without fashion model input
 
     
-    """clip_detected_item = "look" # Default fallback item name
+    clip_detected_item = "look" # Default fallback item name
     clip_detected_item_prob = 0.0
     category_key = 'mid' # Default category
     final_score_str = "N/A"
@@ -324,7 +324,7 @@ def analyze_outfit(input_img):
         
         raw_weighted_score = (drip_score * 1) + (mid_score * 0.5) + (not_score * 0.1)
 
-        final_score = raw_weighted_score * 100
+        final_score = raw_weighted_score * 100 + 10
         
 
         final_score = min(max(final_score, 0), 100)
@@ -369,97 +369,7 @@ def analyze_outfit(input_img):
         
         return ("<p style='color: #FF5555;'>Error during CLIP analysis.</p>",
                 None, f"Analysis Error: {e}")
-    """
-
-    # --- Tunable Scoring Parameters ---
-    # Adjust these weights to control how much each category's average probability contributes
-    # A negative weight for trash means a high trash probability reduces the score.
-    W_DRIP = 100.0
-    W_MID = 40.0 # Reduced weight for mid compared to drip
-    W_TRASH = -60.0 # Negative weight for trash, making it a penalty
-
-    # Adjust these thresholds based on the final scores you observe for different outfits
-    THRESHOLD_DRIP = 65.0 # Score >= this is Drippy
-    THRESHOLD_MID = 30.0 # Score >= this is Mid (and < THRESHOLD_DRIP)
-    # Scores below THRESHOLD_MID will be Trash
-
-    # --- Initialize Variables ---
-    clip_detected_item = "look" # Default fallback item name
-    clip_detected_item_prob = 0.0
-    final_score = 0.0 # Initialize score
-    category_key = 'not_drippy' # Default category before calculation (can be anything, overridden later)
-    score_label = "Score" # Default label before calculation
-
-
-    try:
-        # --- CLIP Processing ---
-        image_tensor = clip_preprocess(cropped_img).unsqueeze(0).to(DEVICE)
-        text_tokens = clip.tokenize(all_prompts).to(DEVICE)
-
-        with torch.no_grad():
-            logits, _ = clip_model(image_tensor, text_tokens)
-            all_probs = logits.softmax(dim=-1).cpu().numpy()[0]
-
-        # --- Calculate Average Probabilities per Category ---
-        drip_len = len(style_prompts['drippy'])
-        mid_len = len(style_prompts['mid'])
-        # Calculate average probabilities for each category
-        drip_avg_prob = np.mean(all_probs[0 : drip_len])
-        mid_avg_prob = np.mean(all_probs[drip_len : drip_len + mid_len])
-        not_avg_prob = np.mean(all_probs[drip_len + mid_len : style_prompts_end_index])
-
-        # --- Calculate Weighted Score (directly scaled to ~0-100) ---
-        # Multiply average probabilities by tunable weights
-        weighted_score = (drip_avg_prob * W_DRIP) + (mid_avg_prob * W_MID) + (not_avg_prob * W_TRASH)
-
-        # --- Final Score & Clipping ---
-        # Clip the score to ensure it's within the 0-100 range
-        final_score = min(max(weighted_score, 0.0), 100.0) # Ensures score is between 0 and 100
-
-        # --- Determine Category based on Tunable Thresholds ---
-        if final_score >= THRESHOLD_DRIP:
-            category_key = 'drippy'
-            score_label = "Drip Score"
-        elif final_score >= THRESHOLD_MID:
-            category_key = 'mid'
-            score_label = "Mid Score"
-        else:
-            category_key = 'not_drippy'
-            score_label = "Trash Score"
-
-        # --- Format Output Strings ---
-        category_label = CATEGORY_LABEL_MAP[category_key]
-        percentage_score_str = f"{final_score:.0f}" # Format as whole number (e.g., "75")
-        # The "/100%" is added in the HTML string below
-
-        # --- Print Debug Info ---
-        print(f"Avg Probs: Drip={drip_avg_prob:.4f}, Mid={mid_avg_prob:.4f}, Trash={not_avg_prob:.4f}")
-        print(f"Weighted Score before clip: {weighted_score:.4f}")
-        print(f"Style analysis: Category={category_label}, Score = {score_label}={percentage_score_str}/100 (Final Score: {final_score:.4f})")
-
-
-        # --- Get top clothing item from CLIP ---
-        # (This part remains largely the same, make sure you copy from here down to the end of the try block)
-        top_3_clip_items = get_top_clip_clothing(all_probs, n=3)
-
-        if top_3_clip_items:
-            detected_items_str = ", ".join([f"{item[0]} ({item[1]*100:.1f}%)" for item in top_3_clip_items]) # Show item and probability
-            print(f"I think I detected: {detected_items_str}")
-            clip_detected_item, clip_detected_item_prob = top_3_clip_items[0]
-        else:
-            print("I couldn't confidently identify specific clothing items via CLIP.")
-            clip_detected_item = "piece" # Use a different fallback if CLIP fails
-            clip_detected_item_prob = 0.0 # Ensure prob is defined
-
-    except Exception as e:
-        print(f"Error during CLIP analysis: {e}")
-        # Keep existing error return format
-        return ("<p style='color: #FF5555;'>Error during CLIP analysis.</p>",
-                None, f"Analysis Error: {e}")
-
-    # --- Determine the Final Item to Mention in Response ---
-    # (This block remains the same after the try block)
-    # ... keep the rest of your analyze_outfit function from here ...
+    
     # Determine the Final Item to Mention in Response
     final_clothing_item = "style" # Ultimate fallback generic term
     generic_response_needed = False