# hkchengrex/XMem

[ECCV 2022] XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model

Repository: https://github.com/hkchengrex/XMem
Canonical: https://ross.abutalabs.com/products/xmem
Homepage: https://hkchengrex.com/XMem/
Language: Python
License: MIT
License Family: permissive
Topics: computer-vision, deep-learning, eccv-2022, eccv2022, pytorch, segmentation, video-object-segmentation, video-segmentation
Last push: 2024-11-15T19:28:38+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 1519, "days_push": 656, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1983, forks 212 (observed 2026-08-28T04:06:01.917630+00:00)

## What it is
XMem is a PyTorch model for semi-supervised video object segmentation that tracks objects through long videos using an Atkinson-Shiffrin-inspired multi-store feature memory. It handles videos of 10,000+ frames at roughly 20 FPS with bounded GPU memory and includes an interactive GUI for annotating frames.

## Use cases
- segment objects in long videos with a mask on the first frame
- track objects through occlusions in video
- annotate video frames with an interactive segmentation GUI
- run video object segmentation on very long clips without running out of GPU memory
- insert masked layers into video footage
- benchmark video object segmentation on DAVIS and long-video datasets

## When to choose
- you need semi-supervised VOS on videos longer than a minute
- GPU memory is a constraint for long-video segmentation
- you want a fast (~20 FPS) segmentation model with an interactive annotation GUI
- you need a research-grade ECCV 2022 baseline for VOS

## When to avoid
- you need fully automatic, prompt-free video segmentation
- you only need image (single-frame) segmentation
- you want a production video-editing tool rather than a research model
- you lack a GPU, as inference speed depends heavily on CUDA hardware

## Facets
- artifact type: library
- maturity: stable
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, deep-learning, machine-learning
- platform: python, cross-platform
- tags: video-object-segmentation, pytorch, eccv-2022, memory-model, semi-supervised-segmentation, gui-demo, video, gpu, linux

## Member repositories
- hkchengrex/XMem (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:01.917630+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 2026-08-30T03:03:26.124359+00:00, confidence not recorded.
  - readme: https://github.com/hkchengrex/XMem (fetched 2026-08-28T04:06:01.917630+00:00, sha 40bc4aa707d2)
  - homepage: https://hkchengrex.com/XMem/ (fetched 2026-08-29T10:43:09.371484+00:00, sha 09fc83b66f4b)
- Data as of 2026-08-30T08:39:29.467469+00:00.