馃敟 #LeAVJEPA: A simple recipe for self-supervised learning from sound and vision 馃敟
One shared encoder, trained without class labels. Dropping a modality makes audio & video learn a common representation.
馃У
One shared encoder, trained without class labels. Dropping a modality makes audio & video learn a common representation.
馃У
0:43