app and requierments fix

app.py

@@ -2,41 +2,26 @@
 import gradio as gr
 import pandas as pd
 import numpy as np
-import glob
 import os
+import io
 from deepface import DeepFace
-from PIL import Image
+from PIL import Image, ImageDraw, ImageFont
 from sklearn.metrics.pairwise import cosine_similarity
 
 # --- 1. Load the Knowledge Base (Specific Target) ---
-# We prioritize your specific file, but keep a fallback just in case
 TARGET_DB = "famous_faces_ArcFace_standalone.parquet"
 
 if os.path.exists(TARGET_DB):
     DB_PATH = TARGET_DB
-else:
-    # Fallback: Find any parquet file if the specific one is missing
-    parquet_files = glob.glob("famous_faces_*.parquet")
-    if parquet_files:
-        # Sort by modification time to get the newest one
-        parquet_files.sort(key=os.path.getmtime, reverse=True)
-        DB_PATH = parquet_files[0]
-    else:
-        DB_PATH = None
-
-if DB_PATH:
     print(f"📂 Loaded Knowledge Base: {DB_PATH}")
     df_db = pd.read_parquet(DB_PATH)
     print(f"📊 Database columns: {df_db.columns.tolist()}")
     print(f"📊 Database shape: {df_db.shape}")
+    
     # Convert embedding column to a clean numpy matrix for fast math
     DB_VECTORS = np.stack(df_db['embedding'].values)
     
     # Identify Model Name from filename
-    # e.g. "famous_faces_GhostFaceNet.parquet" -> "GhostFaceNet"
-    # Identify Model Name from filename
-    # e.g. "famous_faces_GhostFaceNet.parquet" -> "GhostFaceNet"
-    # e.g. "famous_faces_ArcFace_standalone.parquet" -> "ArcFace"
     filename_no_ext = os.path.basename(DB_PATH).replace(".parquet", "")
     parts = filename_no_ext.split("_")
     
@@ -46,96 +31,24 @@ if DB_PATH:
         MODEL_NAME = parts[-1]
     print(f"⚙️ Model configured: {MODEL_NAME}")
     
-    # Debug: Check what directories exist
-    print(f"🔍 Checking available directories...")
-    if os.path.exists("my_dataset"):
-        files_in_dataset = os.listdir("my_dataset")[:5]  # Show first 5
-        print(f"   Found 'my_dataset' directory with {len(os.listdir('my_dataset'))} files")
-        print(f"   Sample files: {files_in_dataset}")
-    else:
-        print(f"   'my_dataset' directory not found")
-    
-    # Debug: Show sample image paths from database
-    if 'image_path' in df_db.columns:
-        sample_paths = df_db['image_path'].head(3).tolist()
-        print(f"📷 Sample image paths from DB: {sample_paths}")
 else:
-    print("❌ CRITICAL: No Parquet file found! Please run the Model Battle step.")
+    print("❌ CRITICAL: Parquet file not found!")
+    DB_PATH = None
     DB_VECTORS = None
     MODEL_NAME = "Unknown"
+    df_db = None
 
-# --- 2. Helper function to load images with fallbacks ---
-def load_image_with_fallback(image_path, name):
-    """
-    Try to load image from various possible locations, or create placeholder.
-    
-    The parquet file stores image_path (e.g., 'my_dataset/Sahar_Milis.png'),
-    but we need to find where the actual image files are located.
-    """
-    tried_paths = []
-    
-    # Strategy 1: Try original path as-is
-    if os.path.exists(image_path):
-        try:
-            return Image.open(image_path), image_path
-        except Exception as e:
-            tried_paths.append(f"{image_path} (error: {e})")
-    else:
-        tried_paths.append(f"{image_path} (not found)")
-    
-    # Strategy 2: Try just the filename in current directory
-    filename = os.path.basename(image_path)
-    if os.path.exists(filename):
-        try:
-            return Image.open(filename), filename
-        except Exception as e:
-            tried_paths.append(f"{filename} (error: {e})")
-    else:
-        tried_paths.append(f"{filename} (not found)")
-    
-    # Strategy 3: Try in my_dataset directory (remove my_dataset/ prefix if present)
-    if image_path.startswith("my_dataset/"):
-        # Already has prefix, try as-is
-        dataset_path = image_path
-    else:
-        # Add prefix
-        dataset_path = os.path.join("my_dataset", filename)
-    
-    if os.path.exists(dataset_path):
-        try:
-            return Image.open(dataset_path), dataset_path
-        except Exception as e:
-            tried_paths.append(f"{dataset_path} (error: {e})")
-    else:
-        tried_paths.append(f"{dataset_path} (not found)")
-    
-    # Strategy 4: Try searching in current directory recursively
-    import glob
-    search_patterns = [
-        f"**/{filename}",
-        f"**/*{filename}",
-        f"**/{name.replace(' ', '_')}*",
-        f"**/{name.replace(' ', '-')}*",
-    ]
-    for pattern in search_patterns:
-        matches = glob.glob(pattern, recursive=True)
-        if matches:
-            try:
-                return Image.open(matches[0]), matches[0]
-            except:
-                continue
-    
-    # Strategy 5: Create a placeholder image with name
+
+# --- 2. Define the Search Logic ---
+def create_placeholder(name):
+    """Creates a placeholder image with the name if the actual image is missing."""
     placeholder = Image.new('RGB', (200, 200), color=(220, 220, 220))
-    from PIL import ImageDraw, ImageFont
     draw = ImageDraw.Draw(placeholder)
-    # Try to use default font, fallback to basic if not available
     try:
         font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 16)
     except:
         font = ImageFont.load_default()
     
-    # Draw name and "Not Found" text
     text_lines = [name[:15], "Image", "Not Found"]
     y_offset = 50
     for line in text_lines:
@@ -144,11 +57,8 @@ def load_image_with_fallback(image_path, name):
         position = ((200 - text_width) // 2, y_offset)
         draw.text(position, line, fill=(100, 100, 100), font=font)
         y_offset += 30
-    
-    print(f"⚠️ Could not find image for {name}. Tried: {tried_paths[:3]}")
-    return placeholder, None
+    return placeholder
 
-# --- 3. Define the Search Logic ---
 def find_best_matches(user_image):
     # Error handling for empty inputs
     if user_image is None: 
@@ -184,63 +94,22 @@ def find_best_matches(user_image):
             display_name = f"{row['name']} (Match: {int(score*100)}%)"
             result_text += f"### #{i}: {display_name}\n\n"
             
-            # Load Image - prioritize bytes from parquet, then try file paths
-            try:
-                img = None
-                found_path = None
-                
-                # Strategy 1: Check if image is stored as bytes in parquet (BEST - self-contained)
-                if 'image_bytes' in df_db.columns and row.get('image_bytes') is not None:
-                    try:
-                        import io
-                        img = Image.open(io.BytesIO(row['image_bytes']))
-                        found_path = "parquet (embedded)"
-                    except Exception as e:
-                        print(f"⚠️ Could not load image bytes for {row['name']}: {e}")
-                
-                # Strategy 2: Try loading from file path (fallback)
-                if img is None and 'image_path' in df_db.columns:
-                    img, found_path = load_image_with_fallback(row['image_path'], row['name'])
-                
-                # Strategy 3: Create placeholder if still no image
-                if img is None:
-                    placeholder = Image.new('RGB', (200, 200), color=(220, 220, 220))
-                    from PIL import ImageDraw, ImageFont
-                    draw = ImageDraw.Draw(placeholder)
-                    try:
-                        font = ImageFont.truetype("/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 16)
-                    except:
-                        font = ImageFont.load_default()
-                    text_lines = [row['name'][:15], "Image", "Not Found"]
-                    y_offset = 50
-                    for line in text_lines:
-                        bbox = draw.textbbox((0, 0), line, font=font)
-                        text_width = bbox[2] - bbox[0]
-                        position = ((200 - text_width) // 2, y_offset)
-                        draw.text(position, line, fill=(100, 100, 100), font=font)
-                        y_offset += 30
-                    img = placeholder
-                    found_path = None
-                
-                gallery_images.append((img, display_name))
-                
-                # Add status message
-                if found_path == "parquet (embedded)":
+            # Load Image from Bytes (Parquet)
+            img = None
+            if 'image_bytes' in df_db.columns and row.get('image_bytes') is not None:
+                try:
+                    img = Image.open(io.BytesIO(row['image_bytes']))
                     result_text += f"✓ Image loaded from parquet\n\n"
-                elif found_path:
-                    result_text += f"✓ Found image at: {found_path}\n\n"
-                else:
-                    result_text += f"⚠️ Image not found for {row['name']}\n\n"
-                    
-            except Exception as img_error:
-                # Even if loading fails, create placeholder
-                placeholder = Image.new('RGB', (200, 200), color=(220, 220, 220))
-                gallery_images.append((placeholder, display_name))
-                result_text += f"⚠️ Error loading image: {str(img_error)}\n\n"
+                except Exception as e:
+                    print(f"⚠️ Could not load image bytes for {row['name']}: {e}")
+            
+            # Fallback to placeholder
+            if img is None:
+                img = create_placeholder(row['name'])
+                result_text += f"⚠️ Image not found (Bytes missing)\n\n"
+                
+            gallery_images.append((img, display_name))
         
-        # Don't pad with None - Gallery can't handle None images
-        # Just return what we have
-
         return gallery_images, result_text
 
     except Exception as e: