En-VidTr-L

Reported on 5 benchmarks across 1 task · 1 paper · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision5 results

VideoonKinetics-700
Top-5 Accuracy· 2021-04-23
89.4
best: 96.7 (UMT-L (ViT-L/16))
SOTA
VidTr: Video Transformer Without Convolutions arXiv:2104.11746
VideoonKinetics-700
Top-1 Accuracy· 2021-04-23
70.8
best: 85.9 (InternVideo2-6B)
VidTr: Video Transformer Without Convolutions arXiv:2104.11746
VideoonCharades
MAP· 2021-04-23
47.3
best: 66.3 (TokenLearner)
VidTr: Video Transformer Without Convolutions arXiv:2104.11746
VideoonKinetics-400
Acc@1· 2021-04-23
80.5
best: 93.6 (OmniVec2)
VidTr: Video Transformer Without Convolutions arXiv:2104.11746
VideoonKinetics-400
Acc@5· 2021-04-23
94.6
best: 98.9 (TubeViT-H (ImageNet-1k))
VidTr: Video Transformer Without Convolutions arXiv:2104.11746