# HobbitLong/PyContrast

PyTorch implementation of Contrastive Learning methods

Repository: https://github.com/HobbitLong/PyContrast
Canonical: https://ross.abutalabs.com/products/pycontrast
Language: Python
License Family: other
Last push: 2023-10-04T01:06:20+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": 2296, "days_push": 1065, "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 1993, forks 180 (observed 2026-08-28T04:06:03.287301+00:00)

## What it is
PyContrast is a PyTorch library providing reference implementations of contrastive learning methods such as InstDis, CMC, and MoCo, along with a curated paper list. It also ships ImageNet unsupervised pre-trained models and tools for transferring them to object detection on PASCAL VOC and COCO.

## Use cases
- implement contrastive learning methods in pytorch
- train self-supervised models like MoCo and CMC
- download imagenet unsupervised pretrained models
- transfer self-supervised pretraining to object detection on VOC and COCO
- find recent contrastive learning papers
- reproduce state-of-the-art self-supervised learning results

## When to choose
- you need reference PyTorch implementations of classic contrastive learning methods
- you want unsupervised ImageNet pre-trained backbones for downstream vision tasks
- you are researching or benchmarking self-supervised representation learning

## When to avoid
- you need the latest self-supervised methods beyond contrastive learning (e.g., masked image modeling)
- you want a maintained general-purpose vision framework rather than research code
- you need a permissively licensed dependency, as the repo has no explicit license

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, computer-vision, deep-learning
- platform: python
- tags: contrastive-learning, self-supervised-learning, pytorch, pretrained-models, image-classification, object-detection

## Member repositories
- HobbitLong/PyContrast (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:03.287301+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 2026-08-30T03:02:32.983510+00:00, confidence not recorded.
  - readme: https://github.com/HobbitLong/PyContrast (fetched 2026-08-28T04:06:03.287301+00:00, sha 60cdf2ada207)
- Data as of 2026-08-30T08:39:29.467469+00:00.