# SHI-Labs/OneFormer

[CVPR 2023] OneFormer: One Transformer to Rule Universal Image Segmentation

Repository: https://github.com/SHI-Labs/OneFormer
Canonical: https://ross.abutalabs.com/products/oneformer
Homepage: https://praeclarumjj3.github.io/oneformer
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: ade20k, cityscapes, coco, image-segmentation, instance-segmentation, panoptic-segmentation, semantic-segmentation, transformer, oneformer, universal-segmentation
Last push: 2024-10-03T03:19:15+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 1400, "days_push": 699, "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 1736, forks 155 (observed 2026-08-28T04:05:29.757972+00:00)

## What it is
OneFormer is a universal image segmentation framework (CVPR 2023) that unifies semantic, instance, and panoptic segmentation in a single transformer model trained once with task tokens and contrastive learning. It provides PyTorch code, pretrained checkpoints, and HuggingFace/Colab integrations for ADE20K, Cityscapes, and COCO.

## Use cases
- run panoptic segmentation on images
- semantic segmentation of street scenes
- instance segmentation on COCO images
- segment images with a single universal model
- fine-tune a segmentation transformer on my dataset
- compare segmentation models on ADE20K benchmarks

## When to choose
- you need one model handling semantic, instance, and panoptic segmentation
- you want state-of-the-art transformer segmentation with pretrained checkpoints
- you want HuggingFace transformers integration for easy inference

## When to avoid
- you need lightweight real-time segmentation on edge devices
- you only need simple classical segmentation like thresholding
- you cannot use GPU resources for training or inference

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, computer-vision, image-processing, deep-learning
- domain: computer-vision, image-processing, deep-learning, artificial-intelligence
- platform: python, cross-platform
- tags: image-segmentation, panoptic-segmentation, semantic-segmentation, instance-segmentation, transformer, pytorch, cvpr-2023, universal-segmentation, huggingface, gpu

## Member repositories
- SHI-Labs/OneFormer (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:29.757972+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 2026-08-30T03:30:40.504330+00:00, confidence not recorded.
  - readme: https://github.com/SHI-Labs/OneFormer (fetched 2026-08-28T04:05:29.757972+00:00, sha b0c325133236)
  - homepage: https://praeclarumjj3.github.io/oneformer (fetched 2026-08-29T11:07:51.853996+00:00, sha 35f5e0419d4c)
- Data as of 2026-08-30T08:39:29.467469+00:00.