File size: 1,767 Bytes
4ee543e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
---
title: "AnyTraverse Studio ๐Ÿšœ"
emoji: "๐Ÿšœ"
colorFrom: "slate"
colorTo: "blue"
sdk: "gradio"
sdk_version: "6.22.0"
app_file: "app.py"
pinned: false
python_version: "3.12"
short_description: "Live off-road traversability evaluation dashboard with Human-in-the-Loop (AnyTraverse)"
tags:
  - computer-vision
  - robotics
  - segmentation
  - vlm
---

# ๐Ÿšœ AnyTraverse Studio โ€” Live Evaluation & HITL Dashboard

A live Gradio dashboard for evaluating the **AnyTraverse** zero-shot off-road
traversability framework ([paper](https://arxiv.org/abs/2506.16826),
[PyPI](https://pypi.org/project/anytraverse/)).

## Workflow
1. **Upload a video** of an off-road scene.
2. Set the traversability preferences (ฯ„), scene-similarity threshold,
   ROI-uncertainty threshold, ROI bounds and the VLM inference **frame skip**.
3. Press **โ–ถ๏ธ Go / Reset**.
4. Watch per frame:
   - raw image + ROI box ยท traversability map ยท uncertainty map ยท ROI crop
   - **attention maps for all prompts** (live, not in the exported video)
   - live dual-metric **ROI traversability + uncertainty** line plot
   - horizontal **0โ€“1 gauge bars** for the two ROI scores
   - the traversal state (`ok` / `unknown_scene` / `unknown_object`)
5. The run **halts** whenever `traversal_state != OK`. Provide an operator
   update like `mud: -0.7; gravel: 0.6` (or just `ok`) and press **Resume**.

Thresholds and ROI bounds can be edited **live** during a run โ€” they are
applied directly to the running pipeline.

## Notes
- The VLM weights (CLIPSeg + CLIP) download on first run and are cached.
- The composed analysis video is exported as an H.264 `.mp4` (bundled
  `imageio-ffmpeg`, no system ffmpeg required).
- GPU is recommended; the space is configured for a **T4** accelerator.