List your company
ResearchFormerly WizWand

Few-shot Video Recognition on Something-Something V2 (K=2, 4, 8, 16)

9.1Top-1 Acc (K=2)Zero-ViCLIP + MSTA + Lcc
2.5484.2495.957.651Nov 18, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Zero-ViCLIP + MSTA + LccGFLOPs=179.84, Transfer learning protocol=Efficient transfer pre-trained models, Multi-view inference=true2024.119.110.413.417.5
Zero-ViCLIP + MSTAGFLOPs=179.84, Transfer learning protocol=Efficient transfer pre-trained models, Multi-view inference=true2024.11910.112.716.6
ViCLIPGFLOPs=161, Transfer learning protocol=Tuning pre-trained models, Multi-view inference=true2024.118.79.711.615.4
Zero-ViCLIP + AdaptFormerGFLOPs=179.80, Transfer learning protocol=Efficient transfer pre-trained models, Multi-view inference=true2024.118.510.111.515.2
Zero-ViCLIP + LORAGFLOPs=179.80, Transfer learning protocol=Efficient transfer pre-trained models, Multi-view inference=true2024.118.49.511.114.9
MOTEGFLOPs=141, Transfer learning protocol=Tuning pre-trained models, Multi-view inference=true2024.117.48.79.812.5
OSTTransfer learning protocol=Tuning pre-trained models, Multi-view inference=true2024.117.38.48.511.5
Vanilla CLIPGFLOPs=282, Transfer learning protocol=Tuning pre-trained models, Multi-view inference=true2024.116.77.3912.2
ViFi-CLIPGFLOPs=281, Transfer learning protocol=Tuning pre-trained models, Multi-view inference=true2024.116.57.78.312.9
Zero-ViCLIPGFLOPs=161, Transfer learning protocol=Efficient transfer pre-trained models, Multi-view inference=true2024.115.15.15.15.1
A5GFLOPs=284, Transfer learning protocol=Adapting pre-trained models, Multi-view inference=true2024.114.55.36.39.9
XCLIPGFLOPs=145, Transfer learning protocol=Adapting pre-trained models, Multi-view inference=true2024.114.45.26.611.1
Vanilla CLIPGFLOPs=281, Transfer learning protocol=Zero-shot pre-trained models, Multi-view inference=true2024.112.82.82.82.8