Action Recognition on ActivityNet (test)
94.3mAPNSNet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| NSNetRecognizer=Swin-L, Pretrain=Kinetics2022.07 | 94.3 | 90.2 | — | |
| TSQNetBackbone=Swin-L, Pretrain=Kinetics2022.07 | 93.7 | 88.7 | — | |
| NSNetRecognizer=Swin-B, Pretrain=Kinetics2022.07 | 91.6 | 86.7 | — | |
| TSQNetBackbone=Swin-B, Pretrain=Kinetics2022.07 | 91.2 | 84.7 | — | |
| MARLRecognizer=SEResNeXt-152, Pretrain=Kinetics2022.07 | 90.1 | 85.7 | — | |
| MARLBackbone=SEResNeXt152, Pretrain=Kinetics2022.07 | 90.1 | — | — | |
| Listen ToLookRecognizer=R(2+1)D-152, Pretrain=Kinetics2022.07 | 89.9 | — | — | |
| ListenToLookBackbone=R(2+1)D-152, Pretrain=Kinetics2022.07 | 89.9 | — | — | |
| NSNetRecognizer=ResNet-152, Pretrain=Kinetics2022.07 | 88.7 | 84.5 | — | |
| DSNRecognizer=R(2+1)D-34, Pretrain=Kinetics2022.07 | 87.8 | 82.6 | — | |
| TSQNetBackbone=ResNet-152, Pretrain=ImageNet2022.07 | 85.2 | 80 | — | |
| NSNetRecognizer=ResNet-152, Pretrain=ImageNet2022.07 | 85.1 | 80.7 | — | |
| Ada3DRecognizer=SlowOnly-50, Pretrain=Kinetics2022.07 | 84.6 | 84 | — | |
| UpperInput frames=16, GFLOPs=117.3, #3D=5.0, #Frame=16.02020.12 | 84.4 | — | — | |
| SMARTRecognizer=ResNet-152, Pretrain=Kinetics2022.07 | 84.4 | — | — | |
| IMGAUD-SKIMMINGBackbone=ResNet-152, Pre-trained=ImageNet2019.12 | 84.2 | 80.3 | — | |
| Listen ToLookRecognizer=ResNet-152, Pretrain=ImageNet2022.07 | 84.2 | 80.3 | — | |
| Ada3DInput frames=16, GFLOPs=70.1, #3D=3.0, #Frame=11.12020.12 | 84 | — | — | |
| MARLBackbone=ResNet-152, Pre-trained=ImageNet2019.12 | 83.8 | 79.8 | — | |
| MARLRecognizer=ResNet-152, Pretrain=ImageNet2022.07 | 83.8 | 79.8 | — | |
| MARLBackbone=ResNet-152, Pretrain=ImageNet2022.07 | 83.8 | 79.8 | — | |
| RRABackbone=ResNet-152, Pre-trained=ImageNet2019.12 | 83.4 | 78.8 | — | |
| RRARecognizer=ResNet-152, Pretrain=ImageNet2022.07 | 83.4 | 78.8 | — | |
| RRABackbone=ResNet-152, Pretrain=ImageNet2022.07 | 83.4 | 78.8 | — | |
| Random FTInput frames=16, GFLOPs=71.3, #3D=3.0, #Frame=11.12020.12 | 82.8 | — | — | |
| UpperInput frames=8, GFLOPs=58.6, #3D=5.0, #Frame=8.02020.12 | 82.6 | — | — | |
| Ada3DInput frames=8, GFLOPs=42.2, #3D=3.1, #Frame=6.62020.12 | 82.6 | — | — | |
| Random FTInput frames=8, GFLOPs=42.9, #3D=3.0, #Frame=6.62020.12 | 81.1 | — | — | |
| RandomInput frames=16, GFLOPs=73.3, #3D=3.0, #Frame=11.22020.12 | 80.4 | — | — | |
| RandomInput frames=8, GFLOPs=42.9, #3D=3.0, #Frame=6.62020.12 | 79.2 | — | — | |
| P3DBackbone=ResNet-152, Pre-trained=ImageNet2019.12 | 78.9 | 75.1 | — | |
| P3DRecognizer=ResNet-152, Pretrain=ImageNet2022.07 | 78.9 | 75.1 | — | |
| P3DBackbone=ResNet-152, Pretrain=ImageNet2022.07 | 78.9 | 75.1 | — | |
| TSQNetFLOPs (G)=26.1, Throughput(bs=1) (videos/s)=17.7, Throughput(bs=32) (videos/s)=121.12022.07 | 76.5 | — | — | |
| FrameExitFLOPs (G)=26.1, Throughput(bs=1) (videos/s)=9.82022.07 | 76.1 | — | — | |
| AdaFocusFLOPs (G)=26.6, Throughput(bs=1) (videos/s)=5.5, Throughput(bs=32) (videos/s)=73.82022.07 | 75 | — | — | |
| IDTPre-trained=ImageNet2019.12 | 68.7 | 64.7 | — | |
| C3DPre-trained=Sports1M2019.12 | 67.7 | 65.8 | — | |
| BIKE ViT-LShot=12022.12 | — | — | 86.6 | |
| BIKE ViT-LShot=22022.12 | — | — | 88.7 | |
| BIKE ViT-LShot=52022.12 | — | — | 90.9 | |
| C3D2017.05 | — | 74.1 | — | |
| Depth2Action2017.05 | — | 78.1 | — | |
| iDT+FV2017.05 | — | 66.5 | — | |
| TSNNumber of segments=32017.05 | — | 89 | — | |
| TSNNumber of segments=72017.05 | — | 89.6 | — | |
| Two Stream2017.05 | — | 71.9 | — |