Update app.py

app.py

@@ -30,6 +30,8 @@ feature_extractors = {model_name: AutoImageProcessor.from_pretrained(model_name)
 classification_models = {model_name: AutoModelForImageClassification.from_pretrained(model_name) for model_name in model_names}
 
 def predict(image, selected_model):
+    if image is None:
+        return None
     feature_extractor = feature_extractors[selected_model]
     model = classification_models[selected_model]
 
@@ -53,19 +55,26 @@ with gr.Blocks() as demo:
     gr.Markdown(description)
     
     with gr.Row():
-        model_dropdown = gr.Dropdown(choices=model_names, label="Selected Model", value=model_names[0])
-        gr.Examples(
-            examples=example_images,
-            inputs=image_input,
-            outputs=output,
-            fn=lambda img: predict(img, model_dropdown.value),
-            cache_examples=True,
-        )
-    image_input = gr.Image(type="pil")
+        model_dropdown = gr.Dropdown(choices=model_names, label="Select Model", value=model_names[0])
+        image_input = gr.Image(type="pil")
     
-    output = gr.Label()
+    # Move the Examples section here, before the output
+    gr.Examples(
+        examples=example_images,
+        inputs=image_input,
+        label="Click on an example image to test",
+    )
+    
+    # Output section
+    output = gr.Label(label="Classification Result")
+
+    # Event handlers
+    def classify(img, model):
+        if img is not None:
+            return predict(img, model)
+        return None
 
-    image_input.change(fn=lambda img: predict(img, model_dropdown.value), inputs=[image_input], outputs=output)
-    model_dropdown.change(fn=lambda img, model: predict(img, model), inputs=[image_input, model_dropdown], outputs=output)
+    image_input.change(fn=classify, inputs=[image_input, model_dropdown], outputs=output)
+    model_dropdown.change(fn=classify, inputs=[image_input, model_dropdown], outputs=output)
 
 demo.launch()