Update app.py

app.py

@@ -1,13 +1,15 @@
 import os
+import io
 import cv2
 import numpy as np
-import requests
 import torch
+import requests
+from PIL import Image
 from ultralytics import YOLO
 import gradio as gr
-from mediapipe import Image as MPImage
 from mediapipe.tasks import python
 from mediapipe.tasks.python import vision
+from mediapipe.tasks.python.vision import Image as MPImage
 import traceback
 
 # -----------------------------
@@ -49,14 +51,28 @@ def predict_asl(image):
         h, w, _ = img.shape
         print(f"🔹 Uploaded image shape: {img.shape}, dtype: {img.dtype}")
 
-        # --- Annotate hand landmarks ---
-        mp_image = MPImage.create_from_array(img)
-        detection_result = detector.detect(mp_image)
-        if detection_result.hand_landmarks:
-            for hand_landmarks in detection_result.hand_landmarks:
-                for landmark in hand_landmarks:
-                    x, y = int(landmark.x * w), int(landmark.y * h)
-                    cv2.circle(img, (x, y), 3, (0, 255, 0), -1)
+        # --- MediaPipe annotation ---
+        try:
+            # Convert OpenCV BGR -> RGB
+            img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+            
+            # PIL + BytesIO to create MediaPipe image
+            pil_img = Image.fromarray(img_rgb)
+            buf = io.BytesIO()
+            pil_img.save(buf, format="PNG")
+            buf.seek(0)
+
+            mp_img = MPImage.create_from_file(buf)
+            
+            detection_result = detector.detect(mp_img)
+            if detection_result.hand_landmarks:
+                for hand_landmarks in detection_result.hand_landmarks:
+                    for landmark in hand_landmarks:
+                        x, y = int(landmark.x * w), int(landmark.y * h)
+                        cv2.circle(img, (x, y), 3, (0, 255, 0), -1)
+        except Exception as e:
+            print("❌ MediaPipe annotation error:", e)
+            traceback.print_exc()
 
         # --- YOLO prediction directly on NumPy array ---
         results = yolo_model.predict(img, imgsz=300, verbose=False)[0]