Action Recognition on ActivityNet
91.2AccuracyHidden Two-Stream Network
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Hidden Two-Stream NetworkBackbone=I3D2017.04 | 91.2 | — | |
| Hidden Two-Stream NetworkBackbone=TSN2017.04 | 87.9 | — | |
| SMARTPre-trained=Kinetics, Backbone=Res152, #F=242020.12 | 84.4 | — | |
| MARLPre-trained=Kinetics, Backbone=Res152, #F=252020.12 | 83.8 | — | |
| RGB DiffFramework=TSN2017.04 | 83 | — | |
| DSNPre-trained=Kinetics, Backbone=Res34, #F=10c2020.12 | 82.6 | — | |
| Hidden Two-Stream NetworkBackbone=VGG162017.04 | 77.8 | — | |
| C3DConfiguration=1 Net2017.04 | 74.1 | — | |
| SMARTPre-trained=Imagenet, Backbone=Res101, #F=102020.12 | 73.1 | — | |
| LiteEvalPre-trained=Imagenet, Backbone=Res1012020.12 | 72.7 | — | |
| AdaframePre-trained=Imagenet, Backbone=Res101, #F=8.652020.12 | 71.5 | — | |
| SMARTPre-trained=Imagenet, Backbone=Res101, #F=82020.12 | 71.4 | — | |
| ActionFlowNet2017.04 | 68.8 | — | |
| DSNPre-trained=Kinetics, Backbone=Res18, #F=10c2020.12 | 68 | — | |
| FrameGlimpsePre-trained=Imagenet, Backbone=VGG16, #F=9.422020.12 | 62.8 | — | |
| FastForwardPre-trained=Imagenet, Backbone=Inc v3, #F=9.612020.12 | 58.1 | — | |
| iCaRL (IM)Buffer Size=40002025.12 | 55.2 | — | |
| iCaRL (IM)Buffer Size=20002025.12 | 50.01 | — | |
| iCaRLBuffer Size=40002025.12 | 46.91 | — | |
| iCaRL (IM)Buffer Size=10002025.12 | 45.76 | — | |
| iCaRLBuffer Size=20002025.12 | 45.05 | — | |
| iCaRLBuffer Size=10002025.12 | 41.72 | — | |
| InternVideo2022.12 | — | 94.3 | |
| InternVideo2_s1-6BTraining Data=IV-2M, Setting=16 x 224, Inference Mode=End-to-end finetuning2024.03 | — | 95.9 | |
| Previous SOTA [1]2022.12 | — | 90.2 | |
| UniFormerV2-LTraining Data=IV-401M, Setting=64 x 336, Inference Mode=End-to-end finetuning2024.03 | — | 94.7 |