gaomingqi/Track-Anything

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

6994 stars · 510 forks observed · 2026-08-28

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

Track-Anything is an interactive tool for video object tracking and segmentation built on Segment Anything, XMem, and E2FGVI. Users specify objects to track via clicks, can correct or change tracked objects during tracking, and use results for downstream tasks like video inpainting and editing.

Use cases

• track and segment objects in videos with click-based interaction
• annotate video object segmentation datasets visually
• segment objects across shot changes in video
• inpaint or remove objects from videos
• correct tracking regions when segmentation drifts
• prepare object masks for video editing workflows

When to choose

• you need interactive, click-based video object segmentation rather than fully automatic pipelines
• you want a visual annotation tool for video segmentation datasets
• you need object masks as input for video inpainting or editing
• your videos contain shot changes requiring re-specification of tracked objects

When to avoid

• you need fully automatic, hands-off video segmentation at scale
• you require real-time tracking performance in production
• you lack a GPU, since the models have significant GPU memory requirements
• you need lightweight integration into an existing application rather than a standalone tool

Facets

application · maturity maintenance

computer-vision image-processing video-processing machine-learning computer-vision image-processing artificial-intelligence python segment-anything video-object-segmentation video-inpainting interactive-annotation gradio xmem e2fgvi video gpu web-server docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="gaomingqi/Track-Anything")