Video Transformer Network

Benchmark Model Rank Results
action-classification-on-moments-in-timeVTN#9Top 1 Accuracy: 37.4Top 5 Accuracy: 65.4
action-classification-on-kinetics-400ViT-B-VTN+ ImageNet-21K (84.0 [10])#100Acc@1: 79.8
action-classification-on-kinetics-400ViT-B-VTN (3 layers, ImageNet pretrain)#117Acc@1: 78.6Acc@5: 93.7
action-classification-on-kinetics-400ViT-B-VTN+ ImageNet-21K (84.0 [10])#191Acc@5: 94.2
action-classification-on-kinetics-400ViT-B-VTN (1 layer, ImageNet pretrain)#193Acc@5: 93.4