CoVeR (JFT-300M)

Reported on 6 benchmarks across 1 task · 1 paper

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

Computer Vision6 results

VideoonKinetics-700
Top-1 Accuracy· uses extra data· 2021-12-14
78.5
best: 85.9 (InternVideo2-6B)
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175
VideoonKinetics-700
Top-5 Accuracy· uses extra data· 2021-12-14
94.2
best: 96.7 (UMT-L (ViT-L/16))
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175
VideoonKinetics-400
Acc@1· 2021-12-14
86.3
best: 93.6 (OmniVec2)
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175
VideoonKinetics-400
Acc@5· 2021-12-14
97.2
best: 98.9 (TubeViT-H (ImageNet-1k))
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175
VideoonKinetics-600
Top-1 Accuracy· uses extra data· 2021-12-14
86.8
best: 91.9 (InternVideo2-6B)
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175
VideoonKinetics-600
Top-5 Accuracy· uses extra data· 2021-12-14
97.3
best: 98.9 (TubeVit-H)
Co-training Transformer with Videos and Images Improves Action Recognition arXiv:2112.07175