I-JEPA

Joint Embedding Predictive Architecture, a self-supervised learning method that predicts abstract representations of masked image regions rather than reconstructing raw pixels.

I-JEPA (Image Joint Embedding Predictive Architecture) is a computer vision architecture developed by Yann LeCun's team that learns visual representations by predicting embeddings of masked image patches. Unlike pixel-reconstruction approaches, I-JEPA operates in an abstract representation space, avoiding the computational cost of predicting every pixel detail. This approach is central to AMI Labs' strategy for building world models that understand physical reality.

Also known as