Update app.py

app.py

@@ -1,27 +1,28 @@
 import gradio as gr
 import numpy as np
-from tensorflow.keras.models import load_model
+from keras.layers import TFSMLayer
 from tensorflow.keras.preprocessing import image
+from PIL import Image
 
-# Load the model from local file (Hugging Face will cache it automatically)
-model = load_model("syaha/skin_cancer_detection_model")
+# Load the model from Hugging Face using TFSMLayer
+model = TFSMLayer("https://huggingface.co/syaha/skin_cancer_detection_model/resolve/main", call_endpoint="serving_default")
 
 # Class names used in the model
 class_names = ['akiec', 'bcc', 'bkl', 'df', 'nv', 'vasc', 'mel']
 
-def predict(img):
+def predict(img: Image.Image):
     img = img.resize((224, 224))
     img_array = image.img_to_array(img)
     img_array = np.expand_dims(img_array, axis=0) / 255.0
 
-    predictions = model.predict(img_array)
+    predictions = model(img_array, training=False).numpy()
     predicted_class = np.argmax(predictions, axis=1)[0]
     label = class_names[predicted_class]
     confidence = float(predictions[0][predicted_class])
 
     return {
         "label": label,
-        "confidence": confidence
+        "confidence": round(confidence * 100, 2)
     }
 
 iface = gr.Interface(
@@ -29,7 +30,7 @@ iface = gr.Interface(
     inputs=gr.Image(type="pil"),
     outputs="json",
     title="Skin Cancer Classifier",
-    description="Upload a skin lesion image to get the predicted class"
+    description="Upload a skin lesion image to get a prediction"
 )
 
 iface.launch()