# HarborYuan/ovsam

[ECCV 2024] The official code of paper "Open-Vocabulary SAM".

Repository: https://github.com/HarborYuan/ovsam
Canonical: https://ross.abutalabs.com/products/ovsam
Homepage: https://www.mmlab-ntu.com/project/ovsam
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-08-04T03:54:26+00:00

## Health v2 (maintenance only)
Score: 42/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 35, release rhythm 35, longevity 69
- inputs: {"age_days": 971, "days_push": 394, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1033, forks 36 (observed 2026-08-28T04:03:18.405796+00:00)

## What it is
Official PyTorch implementation of Open-Vocabulary SAM (ECCV 2024), a model that unifies SAM's interactive segmentation with CLIP's open-vocabulary recognition via SAM2CLIP and CLIP2SAM knowledge transfer modules. It also includes RWKV-SAM, an efficient SAM variant, and offers a Hugging Face demo plus local Gradio demo.

## Use cases
- segment and recognize objects from 22,000 classes interactively
- run open-vocabulary segmentation on custom images
- distill SAM knowledge into CLIP for unified segmentation and recognition
- benchmark open-vocabulary detection on COCO
- try efficient RWKV-based SAM segmentation
- run a local gradio demo of interactive segmentation with recognition

## When to choose
- you need interactive segmentation with open-vocabulary recognition in one model
- you want to reproduce ECCV 2024 Open-Vocabulary SAM or RWKV-SAM research results
- you need lower computational cost than naively combining SAM and CLIP
- you want a pretrained model covering ~22k classes

## When to avoid
- you only need plain SAM segmentation without recognition
- you need a production-ready supported product rather than research code
- you cannot run GPU inference
- you need a non-PyTorch framework

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, computer-vision, image-processing
- domain: computer-vision, image-processing, artificial-intelligence, deep-learning
- platform: python
- tags: segment-anything, clip, open-vocabulary, image-segmentation, zero-shot-recognition, eccv-2024, knowledge-distillation, rwkv, research-code, linux, gpu

## Member repositories
- HarborYuan/ovsam (main) score 42

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:18.405796+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 2026-08-30T07:06:37.351380+00:00, confidence not recorded.
  - readme: https://github.com/HarborYuan/ovsam (fetched 2026-08-28T04:03:18.405796+00:00, sha 473a1bb45a09)
  - homepage: https://www.mmlab-ntu.com/project/ovsam (fetched 2026-08-29T13:06:30.797185+00:00, sha 3ad822d71d8f)
- Data as of 2026-08-30T08:39:29.467469+00:00.