# fudan-zvg/SETR

[CVPR 2021 & IJCV 2024] Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers

Repository: https://github.com/fudan-zvg/SETR
Canonical: https://ross.abutalabs.com/products/setr
Language: Python
License: MIT
License Family: permissive
Last push: 2024-09-02T07:59:01+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2072, "days_push": 730, "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 1108, forks 146 (observed 2026-08-28T04:03:36.775584+00:00)

## What it is
SETR (SEgmentation TRansformers) is the official PyTorch implementation of the CVPR 2021 / IJCV 2024 paper 'Rethinking Semantic Segmentation from a Sequence-to-Sequence Perspective with Transformers'. It provides model configs, pretrained checkpoints, and training/evaluation tooling built on the MMSegmentation framework for pure-transformer semantic segmentation.

## Use cases
- run transformer-based semantic segmentation on cityscapes
- train SETR models on ADE20K and PASCAL Context
- reproduce CVPR 2021 segmentation benchmark results
- use pretrained SETR checkpoints for dense prediction
- compare pure transformer segmentation against CNN baselines
- fine-tune vision transformer segmentation models on custom datasets

## When to choose
- you need a pure-transformer semantic segmentation model with published checkpoints
- you want to reproduce or extend SETR research results
- your stack already uses MMSegmentation/OpenMMLab tooling

## When to avoid
- you need a lightweight segmentation model for production inference on limited hardware
- you want actively developed features or new architectures
- you need a framework-agnostic implementation outside PyTorch/MMSegmentation

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: computer-vision, deep-learning, machine-learning, image-processing
- platform: python
- tags: semantic-segmentation, transformers, vision-transformer, cvpr-2021, mmsegmentation, cityscapes, ade20k, research-code, linux, gpu

## Member repositories
- fudan-zvg/SETR (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:36.775584+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 2026-08-30T06:43:45.947877+00:00, confidence not recorded.
  - readme: https://github.com/fudan-zvg/SETR (fetched 2026-08-28T04:03:36.775584+00:00, sha 214de2fa2071)
- Data as of 2026-08-30T08:39:29.467469+00:00.