add description

README.md

@@ -9,7 +9,7 @@ sdk_version: 5.35.0
 app_file: app.py
 pinned: false
 license: apache-2.0
-short_description: SceneDINO Demo
+short_description: SceneDINO (ICCV 2025)
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -190,6 +190,18 @@ def demo_run(image: str,
     return output_2d, glb_path
 
 
+markdown_description = """
+    [Paper](https://arxiv.org/abs/xxxx.xxxxx) | [Code](https://github.com/tum-vision/scenedino) | [Project Page](https://visinf.github.io/scenedino/)
+    Upload a single image to infer 3D geometry and semantics with **SceneDINO**. You can find some example images below.
+
+    - **Feature PCA**
+    We visualize our high-dimensional feature field using PCA and visualizing three of the components in RGB. Interactively adjust which are visualized.
+    - **SSC** (Semantic Scene Completion)
+    Our features are used downstream for semantic predictions. Choose between the fully unsupervised approach or the linear probing approach.
+
+    <span style="color:orange">⚠️ NOTE: We assume the intrinsic camera matrix of KITTI-360, images are cropped and rescaled to 192x640. Further note our demo's voxel limit of 5M. </span>
+    """
+
 demo = gr.Interface(
     demo_run,
     inputs=[
@@ -197,7 +209,7 @@ demo = gr.Interface(
         gr.Radio(label="Backbone", choices=["DINO (ViT-B)", "DINOv2 (ViT-B)"]),
         gr.Radio(label="Mode", choices=["Feature PCA 1-3", "Feature PCA 4-6", "Feature PCA 7-9", "SSC (unsup.)", "SSC (linear)"]),
         gr.Slider(label="Density threshold", minimum=0, maximum=1, step=0.05, value=0.2),
-        gr.Slider(label="Resolution [m]", minimum=0.05, maximum=0.5, step=0.1, value=0.2),
+        gr.Slider(label="Resolution [m]", minimum=0.05, maximum=0.5, step=0.05, value=0.2),
         gr.Slider(label="X Range [m]", minimum=1, maximum=50, step=1, value=10),
         gr.Slider(label="Y Range [m]", minimum=1, maximum=50, step=1, value=10),
         gr.Slider(label="Z Range [m]", minimum=1, maximum=100, step=1, value=20),
@@ -212,6 +224,7 @@ demo = gr.Interface(
     ],
     title="SceneDINO Demo",
     examples="demo_utils/examples",
+    description=markdown_description,
 )
 
 demo.launch()

demo_utils/examples/log.csv

@@ -1,2 +1,2 @@
 image, mode, sigma_threshold, resolution, x_range, y_range, z_range
-"kitti-360.png","DINO (ViT-B)","Feature PCA 1-3",0.25,0.5,20,20,50
+"kitti-360.png","DINOv2 (ViT-B)","Feature PCA 1-3",0.25,0.25,20,20,50