facebookresearch/vissl

Health v2 · maintenance only

Flags: archived

Adoption not part of the score

3293 stars · 324 forks observed · 2026-08-28

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

VISSL is Facebook AI Research's extensible, modular and scalable PyTorch library for state-of-the-art self-supervised learning with images. It provides reproducible reference implementations of methods like SimCLR, MoCo, SwAV, PIRL and DINO, plus benchmark tasks for evaluating learned representations.

Use cases

• train self-supervised image models like SimCLR or MoCo
• reproduce SOTA self-supervised learning research results
• benchmark pretrained vision representations with linear classification or nearest neighbor tasks
• pretrain vision transformers on large image datasets
• evaluate embeddings for low-shot classification and object detection
• scale self-supervised training across multiple GPUs and nodes

When to choose

• you need reference implementations of many SSL methods in one framework
• you want reproducible benchmarks for evaluating visual representations
• you need multi-GPU/multi-node scaling with FP16 and FSDP support
• you want to build on or extend modular SSL components in PyTorch

When to avoid

• you need supervised-only training without self-supervision features
• you work outside image/vision domains
• you need a lightweight minimal solution rather than a full research framework
• you require frequent updates - the project's latest release was March 2024

Facets

library · maturity maintenance

machine-learning deep-learning image-processing benchmarking computer-vision deep-learning machine-learning python self-supervised-learning pytorch computer-vision representation-learning simclr moco swav dino facebook-research vision-transformers gpu linux docker

Member repositories

For agents

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