# clovaai/CutMix-PyTorch

Official Pytorch implementation of CutMix regularizer

Repository: https://github.com/clovaai/CutMix-PyTorch
Canonical: https://ross.abutalabs.com/products/cutmix-pytorch
Language: Python
License: MIT
License Family: permissive
Topics: cutmix, augmentation, regularization, transfer-learning, iccv2019
Last push: 2020-09-16T03:47:45+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2654, "days_push": 2177, "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 1251, forks 157 (observed 2026-08-28T04:04:08.179469+00:00)

## What it is
Official PyTorch implementation of CutMix, an image data augmentation regularizer that cuts and pastes patches between training images while mixing labels proportionally. It includes training scripts for CIFAR-100 and ImageNet classification and pretrained models.

## Use cases
- augment image classification training data with cutmix
- improve cnn classifier generalization and localization
- train imagenet models with regional dropout alternatives
- get cutmix-pretrained backbones for transfer learning
- reproduce iccv 2019 cutmix paper results

## When to choose
- training image classifiers in PyTorch and wanting the CutMix regularizer
- seeking better robustness and localization than Cutout or Mixup
- needing pretrained models for downstream detection or captioning tasks

## When to avoid
- working outside PyTorch
- needing a maintained library integrated into modern pipelines (repo is research code, last release 2020)
- doing non-image tasks where patch-mixing augmentation does not apply

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, computer-vision, image-processing
- platform: python
- tags: data-augmentation, regularization, image-classification, pytorch, cutmix, research-code, iccv-2019

## Member repositories
- clovaai/CutMix-PyTorch (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:08.179469+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 2026-08-30T05:07:38.726860+00:00, confidence not recorded.
  - readme: https://github.com/clovaai/CutMix-PyTorch (fetched 2026-08-28T04:04:08.179469+00:00, sha 0f90429e25ed)
- Data as of 2026-08-30T08:39:29.467469+00:00.