Update app.py

app.py

@@ -408,8 +408,8 @@ with gr.Blocks(title="REALM Inference Demo") as demo:
                 view2_input = gr.File(label="View 2 (Image, .npz, or .npy)", file_types=["image", ".npz", ".npy"], visible=False)
                 
             run_btn = gr.Button("Run Inference", variant="primary")
+            gr.Markdown("*Note: The depth estimation and semantic segmentation models are specifically evaluated on street-driving scenarios from the MVSEC and DSEC environments. In contrast, the dense feature matching and 3D correspondence pipeline is fully generalizable and works across any arbitrary scene.*", elem_classes="loading-note")
             gr.Markdown("*Note: The interface may pause during the first run of each task while heavy model weights are downloaded from Hugging Face.*", elem_classes="loading-note")
-            
         with gr.Column(scale=1):
             depth_output = gr.Image(label="Predicted Depth", visible=True)
             seg_output = gr.Image(label="Segmentation Mask", visible=False)