Skip to content
Back to skills

Inference And Demo

ASecurity

"Routes BackgroundMattingV2 image, video, webcam, and throughput

  • 247 stars
  • 0 votes
  • 0 copies
  • 2 views
  • Added September 8, 2026
datatestingbackend

Works with

  • cli

Security analysis

A100/100

Pro scans all 6 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill inference-and-demo --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Inference And Demo?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Inference And Demo
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-inference-and-demo-arex-skill/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-inference-and-demo-arex-skill)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: inference-and-demo
description: "Routes BackgroundMattingV2 image, video, webcam, and throughput
  demo workflows."
disable-model-invocation: true
metadata:
  disco-role: operating
license: MIT
---

# inference-and-demo

Use this sub-skill when the task is about running BackgroundMattingV2 demos or
understanding the inference-time model path.

## Typical triggers

- image matting on a directory of source/background images
- video matting on a source video and a background image or video
- webcam matting, background capture, or GUI interaction
- throughput or speed testing on PyTorch or TorchScript
- choosing `mattingbase` vs `mattingrefine`
- choosing `resnet101`, `resnet50`, or `mobilenetv2`
- using `cpu` vs `cuda`
- deciding when `preprocess-alignment` is helpful

## Read first

- `references/workflows.md` for concrete demo recipes and option meaning.
- `references/troubleshooting.md` for missing checkpoints, shape mismatch,
  alignment, webcam, and output-overwrite issues.
- `scripts/smoke_forward.py` for the safe tiny forward smoke.
- `scripts/run_inference_images.py` and `scripts/run_inference_video.py` for
  dry-run or execute wrappers around the checkout's demo CLIs.

## What this sub-skill owns

- model selection and inference-time configuration
- input pairing and alignment behavior
- output types `com`, `pha`, `fgr`, `err`, and `ref`
- small smoke tests that prove the model path works without checkpoints
- guidance for the webcam demo, while acknowledging its hardware and GUI needs

## What it does not own

- TorchScript or ONNX conversion details; use `export-and-backends`.
- Dataset layout and training setup; use `training`.
- Octave/MATLAB benchmark notes; those stay with `training` references.

## Recommended first checks

1. Run `scripts/smoke_forward.py` with a tiny resolution on CPU or CUDA.
2. If you only want to inspect the exact source CLI, use the dry-run wrappers.
3. Move to the native CLI only when you have a real checkpoint and matching
   source/background inputs.

## Cross-links

- `../export-and-backends/SKILL.md`
- `../training/SKILL.md`
- `../../references/backend-compatibility.md`

Files in this skill

  • SKILL.md2.1 KB
  • references/troubleshooting.md1.9 KB
  • references/workflows.md3.7 KB
  • scripts/run_inference_images.py1.7 KB
  • scripts/run_inference_video.py1.7 KB
  • scripts/smoke_forward.py3.8 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.