"Run demo_superglue live matching, headless sequence processing,
Scanned 9/9/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill live-demo-and-visualization --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Live Demo And Visualization?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-live-demo-and-visualization)More formats (shields.io, HTML) on the badges page.
---
name: live-demo-and-visualization
description: "Run demo_superglue live matching, headless sequence processing,
keyboard controls, and visualization troubleshooting."
disable-model-invocation: true
metadata:
disco-role: operating
license: NOASSERTION
---
# Live Demo and Visualization
Use this sub-skill for `demo_superglue.py`-style live matching when the goal is to watch matches update against an anchor frame, write output images, or run safely on a remote server without a GUI.
## Route elsewhere
- Batch image-pair matching, `.npz` dumps, pose evaluation, or pair-file validation: [`../pair-matching-evaluation/`](../pair-matching-evaluation/)
- Direct Python API work with `models.Matching`, `SuperPoint`, `SuperGlue`, or tensor-level inference: [`../programmatic-api/`](../programmatic-api/)
## Use the bundled references
- [CLI reference](references/cli-reference.md)
- [Workflows](references/workflows.md)
- [Troubleshooting](references/troubleshooting.md)
## Bundled helper
- [`scripts/run_headless_demo_smoke.py`](scripts/run_headless_demo_smoke.py): bounded directory-mode smoke helper that always uses `--no_display` and forces CPU when requested.
## Operating notes
- `demo_superglue.py` uses `VideoStreamer` to normalize webcam, IP/RTSP, video-file, and directory inputs.
- The first frame becomes the initial anchor; press `n` to replace it with the current frame.
- `make_matching_plot_fast` renders confidence-colored lines and optional keypoints; `k` toggles keypoints in the interactive viewer.
- Prefer `--no_display` plus `--output_dir` for remote or headless runs.
- When few or no matches appear, first check resize, model choice, thresholds, and the selected anchor frame.
- Keep this sub-skill focused on visualization and interactive demo flows; do not use it for pair evaluation or Python API guidance.Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!