Update model card with cleaner examples

README.md

@@ -27,15 +27,15 @@ results = model.predict('image.jpg')
 
 # Get the prediction
 probs = results[0].probs
-class_id = probs.top1  # 0=hand, 1=arm, 2=not_hand
+class_id = probs.top1  # 0=arm, 1=hand, 2=not_hand (alphabetical order!)
 confidence = probs.top1conf.item()
 
 # Interpret results
-if class_id == 0:
+if class_id == 1:  # hand is index 1
     print(f"✋ Hand detected: {confidence:.1%}")
-elif class_id == 1:
+elif class_id == 0:  # arm is index 0
     print(f"💪 Arm detected: {confidence:.1%}")
-else:
+else:  # not_hand is index 2
     print(f"❌ No hand/arm detected: {confidence:.1%}")
 ```
 
@@ -56,7 +56,8 @@ while True:
     results = model(frame)
     probs = results[0].probs
 
-    classes = ['hand', 'arm', 'not_hand']
+    # YOLO uses alphabetical order!
+    classes = ['arm', 'hand', 'not_hand']  # 0=arm, 1=hand, 2=not_hand
     label = f"{classes[probs.top1]}: {probs.top1conf:.1%}"
 
     cv2.putText(frame, label, (10, 30),
@@ -163,7 +164,7 @@ async def detect(file: UploadFile = File(...)):
     probs = results[0].probs
 
     return {
-        "class": ['hand', 'arm', 'not_hand'][probs.top1],
+        "class": ['arm', 'hand', 'not_hand'][probs.top1],  # alphabetical order
         "confidence": float(probs.top1conf)
     }
 ```
@@ -231,7 +232,7 @@ curl -X POST -F "file=@test.jpg" http://localhost:8000/detect
 ## Model Details
 
 - **Architecture**: YOLOv8s-cls (5M parameters)
-- **Classes**: 3 (hand, arm, not_hand)
+- **Classes**: 3 (arm=0, hand=1, not_hand=2) - alphabetical order
 - **Input Size**: 224x224
 - **Accuracy**: >96% on validation set
 - **Size**: ~3MB