Fetching the paper…

VIMPAC: Video Pre-Training via Masked Token Prediction and Contrastive Learning · Around