Update app.py

app.py

@@ -460,25 +460,125 @@ class BrandCatalog:
             ranked = [(i, sims[i]) for i in np.argsort(sims)[::-1][:top_k]]
         return [dict(self.items[i], score=float(s), source='Catalog') for i, s in ranked]
 
+# ============================================
+# Personal Wardrobe (User Upload)
+# ============================================
+
+class PersonalWardrobe:
+    """
+    Manages user's personal wardrobe - uploaded images in session
+    Items stored in memory (not persistent across sessions)
+    """
+    def __init__(self, pipeline, color_engine):
+        self.pipe = pipeline
+        self.color_engine = color_engine
+        self.items = []
+        self.embeddings = None
+    
+    def add_item(self, image_path, category='Unknown', name=None):
+        """Add a clothing item to personal wardrobe"""
+        if not os.path.exists(image_path):
+            return None
+        
+        try:
+            # Extract features
+            features = self.pipe.extractor.extract_all_features(image_path, category)
+            
+            # Compute embedding
+            emb = self.pipe.embed(image_path)
+            
+            item = {
+                'id': len(self.items),
+                'name': name or f"My {category} #{len(self.items)+1}",
+                'image_path': image_path,
+                'category': category,
+                'colors': features['colors'],
+                'style': features['style'],
+                'fabric': features['fabric'],
+                'texture': features.get('texture', 'Solid'),
+            }
+            
+            self.items.append(item)
+            
+            # Add embedding
+            if self.embeddings is None:
+                self.embeddings = emb.reshape(1, -1)
+            else:
+                self.embeddings = np.vstack([self.embeddings, emb.reshape(1, -1)])
+            
+            return item
+        except Exception as e:
+            print(f"Error adding item: {e}")
+            return None
+    
+    def remove_item(self, item_id):
+        """Remove item from wardrobe by ID"""
+        if 0 <= item_id < len(self.items):
+            self.items.pop(item_id)
+            if self.embeddings is not None:
+                self.embeddings = np.delete(self.embeddings, item_id, axis=0)
+                if len(self.embeddings) == 0:
+                    self.embeddings = None
+            # Re-index remaining items
+            for i, it in enumerate(self.items):
+                it['id'] = i
+            return True
+        return False
+    
+    def get_all_items(self):
+        """Get all wardrobe items"""
+        return self.items
+    
+    def clear_all(self):
+        """Clear entire wardrobe"""
+        self.items = []
+        self.embeddings = None
+    
+    def search(self, query_emb, top_k=5, cat_filter=None):
+        """Search wardrobe for matching items"""
+        if self.embeddings is None or not self.items:
+            return []
+        
+        sims = cosine_similarity([query_emb], self.embeddings)[0]
+        
+        if cat_filter:
+            cf_list = cat_filter if isinstance(cat_filter, list) else [cat_filter]
+            idxs = [i for i, it in enumerate(self.items)
+                    if any(cf.lower() in it.get('category', '').lower() for cf in cf_list)]
+            if idxs:
+                ranked = sorted([(i, sims[i]) for i in idxs],
+                               key=lambda x: x[1], reverse=True)[:top_k]
+            else:
+                ranked = [(i, sims[i]) for i in np.argsort(sims)[::-1][:top_k]]
+        else:
+            ranked = [(i, sims[i]) for i in np.argsort(sims)[::-1][:top_k]]
+        
+        return [dict(self.items[i], score=float(s), source='My Wardrobe') for i, s in ranked]
+
 # ============================================
 # Outfit Recommender
 # ============================================
 
 class OutfitRecommender:
-    def __init__(self, pipeline, color_engine, catalog):
+    def __init__(self, pipeline, color_engine, catalog, personal_wardrobe=None):
         self.pipe = pipeline
         self.ce = color_engine
         self.cat = catalog
+        self.pw = personal_wardrobe  # Personal Wardrobe
     
     def recommend(self, image_path, top_k=5, style='Casual', occasion='everyday',
-                  complement=True, query_category=None):
+                  complement=True, query_category=None, use_personal_wardrobe=False):
         qf = self.pipe.extractor.extract_all_features(image_path, query_category or 'query')
         qe = self.pipe.embed(image_path)
         
         cat_filter = _get_complement_cats(query_category or '') if complement else None
         
         results = []
-        if self.cat and self.cat.items:
+        
+        # Choose source: Personal Wardrobe or Catalog
+        if use_personal_wardrobe and self.pw and self.pw.items:
+            results.extend(self.pw.search(qe, top_k * 2, cat_filter))
+        elif self.cat and self.cat.items:
             results.extend(self.cat.search(qe, top_k * 2, cat_filter))
         
         for r in results:
@@ -632,7 +732,11 @@ print(f"   ❌ Skipped - no image path: {skipped_no_path}")
 print(f"   ❌ Skipped - file not found: {skipped_no_file}")
 print(f"   ❌ Skipped - processing error: {processed_error}")
 
-outfit_recommender = OutfitRecommender(pipeline, color_engine, brand_catalog)
+# Initialize Personal Wardrobe
+personal_wardrobe = PersonalWardrobe(pipeline, color_engine)
+print(f"\n👤 Personal Wardrobe initialized (0 items)")
+
+outfit_recommender = OutfitRecommender(pipeline, color_engine, brand_catalog, personal_wardrobe)
 
 if len(brand_catalog.items) > 0:
     print(f"\n✅ System ready with {len(brand_catalog.items)} clothing items!")
@@ -651,17 +755,45 @@ def _save_temp(img_array, name='query'):
     Image.fromarray(img_array).save(path, quality=95)
     return path
 
-def fn_recommend(image, style, occasion, complement, query_cat):
+def fn_recommend(image, style, occasion, complement, query_cat, use_personal_wardrobe=False):
     if image is None:
         return [], "⚠️ Please upload an image."
     
+    # Check which source to use
+    if use_personal_wardrobe:
+        if len(personal_wardrobe.items) == 0:
+            return [], "⚠️ Your personal wardrobe is empty! Upload items first in the 'My Wardrobe' tab."
+    else:
+        if len(brand_catalog.items) == 0:
+            error_msg = """❌ No items in catalog!
+            
+The system has 0 items because image files are missing.
+
+🔧 To fix this:
+1. Check that images exist at paths in features_metadata.csv
+2. Current paths look like: /kaggle/working/preprocessed/...
+3. Upload images to Hugging Face Space in 'images' folder
+4. Run fix_image_paths.py to update CSV paths
+5. Replace features_metadata.csv with the fixed version
+
+📁 Expected structure:
+   images/
+   ├── tops/
+   ├── bottoms/
+   ├── outerwear/
+   └── dresses/
+            """
+            return [], error_msg
+    
     tmp = _save_temp(image)
     res = outfit_recommender.recommend(
         tmp, top_k=6, style=style, occasion=occasion,
-        complement=complement, query_category=query_cat if query_cat != 'Auto' else None)
+        complement=complement, query_category=query_cat if query_cat != 'Auto' else None,
+        use_personal_wardrobe=use_personal_wardrobe)
     
     gallery = []
-    lines = [f"🔍 Detected: Colors={', '.join(res['query_features'].get('colors',[])[:3])} | Style={res['query_features'].get('style','N/A')}"]
+    source_label = "🏠 My Wardrobe" if use_personal_wardrobe else "🏬 Brand Catalog"
+    lines = [f"{source_label} | 🔍 Detected: Colors={', '.join(res['query_features'].get('colors',[])[:3])} | Style={res['query_features'].get('style','N/A')}"]
     
     _qcat = query_cat if query_cat != 'Auto' else 'Auto-detected'
     _comp_cats = _get_complement_cats(query_cat if query_cat != 'Auto' else '')