facebookresearch/ijepa

Health v2 · maintenance only

Flags: no_releases archived no_license

Adoption not part of the score

3489 stars · 522 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Official PyTorch implementation of I-JEPA, a self-supervised learning method that predicts latent representations of image regions from other regions without hand-crafted augmentations or pixel-level reconstruction. It provides training and evaluation code for pretraining vision encoders that learn strong semantic representations.

Use cases

• pretrain a vision encoder with self-supervised learning
• learn image representations without data augmentation
• reproduce I-JEPA CVPR 2023 paper results
• evaluate pretrained features on image classification
• research joint-embedding predictive architectures
• train efficient SSL models on limited compute

When to choose

• you need strong off-the-shelf semantic image features without hand-crafted augmentations
• you want a compute-efficient self-supervised pretraining method
• you are researching latent-space predictive architectures

When to avoid

• you need a production-ready pretrained model zoo with many checkpoints
• you want generative pixel-level image modeling
• you need non-PyTorch framework support

Facets

library · maturity active

machine-learning deep-learning machine-learning deep-learning computer-vision image-processing python self-supervised-learning pytorch computer-vision research-code joint-embedding cvpr gpu

Member repositories

For agents

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