Update app.py

app.py

@@ -160,58 +160,31 @@ def preprocess_image(image_data):
     - Convert to numpy array, add batch dim
     """
     try:
-        # Load image
-        image = Image.open(io.BytesIO(image_data)).convert('RGB')
-        w, h = image.size  # (width=400, height=296 expected)
-
-        # # Define ROI (x1:x2, y1:y2), inclusive indices
-        # X1, X2 = 90, 280
-        # Y1, Y2 = 5, 205
-
-        # # Clamp to actual image bounds
-        # X1 = max(0, min(w - 1, X1))
-        # X2 = max(0, min(w - 1, X2))
-        # Y1 = max(0, min(h - 1, Y1))
-        # Y2 = max(0, min(h - 1, Y2))
-
-        # # Pillow crop box is (left, upper, right, lower) with right/lower EXCLUSIVE
-        # crop_box = (X1, Y1, X2 + 1, Y2 + 1)
-        # image = image.crop(crop_box)
+        # Convert bytes to PIL Image
+        image = Image.open(io.BytesIO(image_data)).convert("RGB")
+        
+        # Crop (x1=450, y1=400, x2=1090, y2=1060)
+        crop_box = (450, 400, 1090, 1060)  
+        image = image.crop(crop_box)
 
         # Resize to model input size
-        image = image.resize((224, 224), Image.BICUBIC)
+        image = image.resize((224, 224), Image.Resampling.LANCZOS)
 
-        # Convert to numpy array
-        image_array = np.array(image, dtype=np.float32)
+        # Normalize and expand dims
+        image = np.array(image).astype("float32")
 
         # Add batch dimension
-        image_array = np.expand_dims(image_array, axis=0)
+        image = np.expand_dims(image, axis=0)
+
 
         # Model has Rescaling(1./255) layer, so no manual normalization
-        return image_array
+        return image
 
     except Exception as e:
         logger.error(f"Image preprocessing error: {e}")
         raise HTTPException(status_code=400, detail=f"Image preprocessing failed: {e}")
 
 
-@app.on_event("startup")
-async def startup_event():
-    """Load model on startup"""
-    global model
-    try:
-        model = load_model()
-        logger.info("API startup complete")
-        
-        # Test model with dummy input
-        dummy_input = np.random.random((1, 224, 224, 3)).astype(np.float32)
-        _ = model.predict(dummy_input, verbose=0)
-        logger.info("Model test prediction successful")
-        
-    except Exception as e:
-        logger.error(f"Startup failed: {e}")
-        raise
-
 @app.get("/health")
 async def health_check():
     """Health check endpoint"""