facebookresearch/simsiam

Health v2 · maintenance only

Flags: no_releases archived no_license

Adoption not part of the score

1237 stars · 174 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A PyTorch implementation of SimSiam, a simple Siamese-based self-supervised representation learning method from Facebook AI Research (CVPR 2021). It provides training scripts for unsupervised pre-training of ResNet-50 on ImageNet and linear classification evaluation, with pre-trained checkpoints.

Use cases

• train a self-supervised image representation model on ImageNet
• reproduce SimSiam paper results
• pre-train a ResNet-50 backbone without labels
• evaluate learned features with linear classification
• compare self-supervised learning methods like MoCo and SimSiam

When to choose

• you need a reference implementation of SimSiam for research or reproduction
• you want to pre-train vision backbones without labels on multi-GPU hardware
• you need released pre-trained checkpoints for transfer learning experiments

When to avoid

• you need single-GPU or DataParallel training - only DistributedDataParallel is supported
• you want a maintained library with new features - the repo is a research code release
• you need a plug-and-play self-supervised framework rather than paper-specific scripts

Facets

library · maturity maintenance

machine-learning deep-learning machine-learning computer-vision deep-learning python self-supervised-learning representation-learning contrastive-learning pytorch research-code imagenet resnet gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="facebookresearch/simsiam")