Update app.py

app.py

@@ -8,15 +8,15 @@ def validate(item_id: str, threshold: float):
     images, domain, probas, valid_probas, invalid_probas = detector.predict_probas_item(item_id)
     valid_images = [x for i, x in enumerate(images) if valid_probas[i].squeeze() >= threshold]
     invalid_images = [x for i, x in enumerate(images) if valid_probas[i].squeeze() < threshold]
-    return f"<h2>Domain: {domain}</h2>", valid_images, invalid_images
+    return f"## Domain: {domain}", valid_images, invalid_images
 
 with gr.Blocks() as demo:
     gr.Markdown("""
                 # Off topic image detector
-                ## This app takes an item ID and classifies its pictures as valid/invalid depending on whether they relate to the domain in which it's been listed.
+                ### This app takes an item ID and classifies its pictures as valid/invalid depending on whether they relate to the domain in which it's been listed.
                 Input an item ID or select one of the preloaded examples below.""")
     item_id = gr.Textbox(label="Item ID")
-    threshold = gr.Number(label="Threshold", value=0.5)
+    threshold = gr.Number(label="Threshold", value=0.25)
     submit = gr.Button("Submit")
     gr.HTML("<hr>")
     domain = gr.Markdown()
@@ -26,7 +26,7 @@ with gr.Blocks() as demo:
     submit.click(inputs=[item_id, threshold], outputs=[domain, valid, invalid], fn=validate)
     gr.HTML("<hr>")
     gr.Examples(
-        examples=[["MLU449951849", 0.3], ["MLA1293465558", 0.3], ["MLB3184663685", 0.3], ["MLC1392230619", 0.3]],
+        examples=[["MLU449951849", 0.25], ["MLA1293465558", 0.25], ["MLB3184663685", 0.25], ["MLC1392230619", 0.25]],
         inputs=[item_id, threshold],
         outputs=[domain, valid, invalid],
         fn=validate,