Update app.py

app.py

@@ -56,6 +56,7 @@ def siglip_detector(image, texts):
     outputs = siglip_model(**inputs)
     logits_per_image = outputs.logits_per_image
     probs = torch.sigmoid(logits_per_image)
+    probs = normalize_tensor(probs)
     
   return probs
 
@@ -81,7 +82,7 @@ def infer(image, labels):
 
 with gr.Blocks() as demo:
   gr.Markdown("# Compare Multilingual Zero-shot Image Classification")
-  gr.Markdown("Compare the performance of SigLIP and othe rmodels on zero-shot classification in this Space 👇")
+  gr.Markdown("Compare the performance of SigLIP and othe rmodels on zero-shot classification in this Space. Three models are compared: CLIP-ViT, NLLB-CLIP and SigLIP-Multilingual. Note that SigLIP outputs are normalized for visualization purposes.")
   with gr.Row():
     with gr.Column():
         image_input = gr.Image(type="pil")