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

Benchmark Model Rank Results
action-recognition-on-diving-48VIMPAC#8Accuracy: 85.5
action-recognition-in-videos-on-ucf101VIMPAC#393-fold Accuracy: 92.7
action-recognition-in-videos-on-hmdb-51VIMPAC#41Average accuracy of 3 splits: 65.9
action-recognition-in-videos-on-something-something-v2VIMPAC#50Top-1 Accuracy: 68.1
action-classification-on-kinetics-400VIMPAC#135Acc@1: 77.4