senguptaumd/Background-Matting

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

4769 stars · 653 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Official research code for 'Background Matting: The World is Your Green Screen' (CVPR 2020), a deep network that extracts per-pixel alpha mattes and foreground color of people from photos and videos using an extra background-only image instead of a green screen. It includes inference code for images and videos, training code on synthetic composites and unlabeled real videos, and a captured dataset.

Use cases

• remove background from portrait photos without a green screen
• extract a person from video and composite onto a new background
• generate alpha mattes for video calls or virtual backgrounds
• train a matting model on synthetic composite data
• self-supervised matting training on unlabeled real videos

When to choose

• you need research-grade background matting with an extra background frame available
• you want to reproduce CVPR 2020 matting results or build on the training code
• you need per-pixel alpha mattes rather than coarse segmentation masks

When to avoid

• you need real-time 4K performance - use BackgroundMattingV2 instead
• you cannot capture a background-only image of the scene
• you need a permissively licensed production dependency - it is CC BY-NC-SA and unmaintained

Facets

library · maturity maintenance

machine-learning image-processing computer-vision computer-vision image-processing deep-learning python background-matting alpha-matting green-screen video-segmentation research-code cvpr-2020 linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="senguptaumd/Background-Matting")