# SysCV/sam-pt

SAM-PT: Extending SAM to zero-shot video segmentation with point-based tracking.

Repository: https://github.com/SysCV/sam-pt
Canonical: https://ross.abutalabs.com/products/sam-pt
Homepage: https://arxiv.org/abs/2307.01197
Language: Python
License: Apache-2.0
License Family: permissive
Topics: segment-anything, segment-anything-model, video-instance-segmentation, video-object-segmentation, zero-shot-segmentation, interactive-video-segmentation, track-anything
Last push: 2024-01-27T23:10:56+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 83
- inputs: {"age_days": 1164, "days_push": 949, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1041, forks 64 (observed 2026-08-28T04:03:20.545062+00:00)

## What it is
SAM-PT extends the Segment Anything Model to zero-shot video segmentation by combining SAM with sparse point-based tracking (PIPS, CoTracker). Users annotate a few points on the first frame and interactively refine masks across video.

## Use cases
- segment objects in video from a few point clicks
- zero-shot video object segmentation without training
- interactively refine video segmentation masks by adding or removing points
- evaluate segmentation on DAVIS, YouTube-VOS, or UVO benchmarks
- track and segment multiple objects in open-world videos

## When to choose
- you need zero-shot video segmentation without training a model
- you want interactive point-based annotation for video objects
- you need a research baseline combining SAM with point trackers

## When to avoid
- you need production-grade, actively maintained video segmentation
- you prefer mask-propagation methods like XMem over point tracking
- you lack a GPU or cannot download large model checkpoints

## Facets
- artifact type: library
- maturity: maintenance
- function: computer-vision, image-processing, machine-learning
- domain: computer-vision, deep-learning
- platform: python
- tags: segment-anything, video-object-segmentation, point-tracking, zero-shot-segmentation, interactive-segmentation, sam, video, research, gpu, linux

## Member repositories
- SysCV/sam-pt (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:20.545062+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:02:52.734483+00:00, confidence not recorded.
  - readme: https://github.com/SysCV/sam-pt (fetched 2026-08-28T04:03:20.545062+00:00, sha 1365e909c83e)
  - homepage: https://arxiv.org/abs/2307.01197 (fetched 2026-08-29T13:04:15.812000+00:00, sha 228aa7220498)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T13:04:15.821906+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T13:04:15.890595+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T13:04:15.892596+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T13:04:15.884719+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.