# RoyalVane/CLAN

( TPAMI2022 / CVPR2019 Oral ) Taking A Closer Look at Domain Shift: Category-level Adversaries for Semantics Consistent Domain Adaptation

Repository: https://github.com/RoyalVane/CLAN
Canonical: https://ross.abutalabs.com/products/clan
Language: Python
License: MIT
License Family: permissive
Topics: semantic-segmentation, domain-adaptation, adversarial-learning, transfer-learning
Last push: 2021-03-13T03:45:41+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": 2711, "days_push": 1999, "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 1289, forks 44 (observed 2026-08-28T04:04:15.206286+00:00)

## What it is
Official PyTorch implementation of CLAN, a CVPR 2019 (oral) / TPAMI 2022 method for unsupervised domain adaptation in semantic segmentation using category-level adversaries with semantics-consistent alignment. It includes training, evaluation, and IoU computation scripts for adapting segmentation models from synthetic datasets (GTA5, SYNTHIA) to the real-world Cityscapes dataset, plus pretrained model checkpoints.

## Use cases
- adapt a semantic segmentation model from synthetic images to real street scenes
- reproduce CLAN CVPR 2019 results on GTA5-to-Cityscapes
- benchmark my segmentation model against a domain adaptation baseline
- run unsupervised domain adaptation without source labels at test time
- compute mIoU for segmentation predictions on Cityscapes validation
- evaluate many training checkpoints in bulk to find the best epoch

## When to choose
- you need the exact reference implementation for a domain adaptation paper or comparison baseline
- you are doing research on adversarial alignment for semantic segmentation and want a starting codebase
- you need pretrained GTA5/SYNTHIA-to-Cityscapes segmentation models to evaluate or build on

## When to avoid
- you need a maintained, general-purpose domain adaptation library for production workloads
- you require a modern PyTorch version, extensive docs, or APIs beyond the paper's experiments
- your task is not semantic segmentation or you lack a GPU with at least 11GB memory

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, computer-vision, image-processing
- domain: computer-vision, machine-learning, deep-learning, image-processing
- platform: python
- tags: semantic-segmentation, domain-adaptation, adversarial-learning, transfer-learning, pytorch, deep-learning-research, gta5, synthia, cityscapes, scene-understanding, pretrained-models, research-code, gpu, linux

## Member repositories
- RoyalVane/CLAN (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:15.206286+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 2026-08-30T04:55:42.304727+00:00, confidence not recorded.
  - readme: https://github.com/RoyalVane/CLAN (fetched 2026-08-28T04:04:15.206286+00:00, sha 085711aebe3b)
- Data as of 2026-08-30T08:39:29.467469+00:00.