Action Recognition on Jester (val)
97.4Top-1 AccuracyPAN_En
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PAN_EnBackbone=ResNet-101+TSM, Flow=false, #Frame=(8+8x4)x2, FLOPs x views=(85.6G+166.1G)x22020.08 | 97.4 | 99.9 | — | — | — | |
| PAN_EnBackbone=ResNet-50+TSM, Flow=false, #Frame=(8+8x4)x2, FLOPs x views=(46.6G+88.4G)x22020.08 | 97.2 | 99.9 | — | — | — | |
| STMBackbone=ResNet-50, Frame=162019.08 | 96.7 | 99.9 | — | — | — | |
| MFNet-C101K (number of training segments)=102018.07 | 96.68 | 99.84 | — | — | — | |
| STMBackbone=ResNet-50, Frame=82019.08 | 96.6 | 99.9 | — | — | — | |
| PAN_FullBackbone=ResNet-50+TSM, Flow=false, #Frame=8+8x4, FLOPs x views=67.7Gx12020.08 | 96.6 | 99.8 | — | — | — | |
| MFNet-C50K (number of training segments)=102018.07 | 96.56 | 99.82 | — | — | — | |
| MFNet-S50K (number of training segments)=102018.07 | 96.5 | 99.86 | — | — | — | |
| MFNet-S50K (number of training segments)=72018.07 | 96.31 | 99.8 | — | — | — | |
| PAN_LiteBackbone=ResNet-50+TSM, Flow=false, #Frame=8+8x4, FLOPs x views=35.7Gx12020.08 | 96.2 | 99.8 | — | — | — | |
| MFNet-C50K (number of training segments)=72018.07 | 96.13 | 99.65 | — | — | — | |
| MFNet-C50Backbone=ResNet-50, Frame=72019.08 | 96.1 | 99.7 | — | — | — | |
| MultiScale TRNevaluation_strategy=10-crop2018.07 | 95.31 | 99.86 | — | — | — | |
| TRN-MultiscaleBackbone=BNInception, Frame=82019.08 | 95.3 | — | — | — | — | |
| TSMBackbone=ResNet-50, Frame=162019.08 | 95.3 | 99.8 | — | — | — | |
| TRN-MultiscaleBackbone=BNInception, Flow=false, #Frame=8, FLOPs x views=16Gx12020.08 | 95.3 | — | — | — | — | |
| TSM_16FBackbone=ResNet-50, Flow=false, #Frame=16, FLOPs x views=65Gx12020.08 | 95.3 | 99.8 | — | — | — | |
| ResNeXt-101Batch Size=8, Input Frames=16, Resolution=112x1122019.04 | 94.89 | — | — | — | — | |
| 3D-MobileNetV2 1.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 94.59 | — | — | — | — | |
| TSMBackbone=ResNet-50, Frame=82019.08 | 94.4 | 99.7 | — | — | — | |
| TSM_8FBackbone=ResNet-50, Flow=false, #Frame=8, FLOPs x views=33Gx12020.08 | 94.4 | 99.7 | — | — | — | |
| ResNet-101Batch Size=8, Input Frames=16, Resolution=112x1122019.04 | 94.1 | — | — | — | — | |
| 3D-ShuffleNetV2 2.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.71 | — | — | — | — | |
| MultiScale TRN2018.07 | 93.7 | 99.59 | — | — | — | |
| ResNet-50Batch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.7 | — | — | — | — | |
| 3D-ShuffleNetV1 2.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.54 | — | — | — | — | |
| 3D-MobileNetV2 0.7xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.34 | — | — | — | — | |
| ResNet-18Batch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.34 | — | — | — | — | |
| 3D-ShuffleNetV2 1.5xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.16 | — | — | — | — | |
| 3D-ShuffleNetV1 1.5xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 93.12 | — | — | — | — | |
| 3D-MobileNetV1 2.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 92.56 | — | — | — | — | |
| 3D-ShuffleNetV1 1.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 92.27 | — | — | — | — | |
| 3D-ShuffleNetV2 1.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 91.96 | — | — | — | — | |
| 3D-MobileNetV1 1.5xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 91.28 | — | — | — | — | |
| 3D-MobileNetV1 1.0xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 90.81 | — | — | — | — | |
| 3D-SqueezeNetBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 90.77 | — | — | — | — | |
| 3D-MobileNetV2 0.45xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 90.21 | — | — | — | — | |
| 3D-ShuffleNetV1 0.5xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 89.23 | — | — | — | — | |
| 3D-MobileNetV1 0.5xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 87.61 | — | — | — | — | |
| 3D-ShuffleNetV2 0.25xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 86.91 | — | — | — | — | |
| 3D-MobileNetV2 0.2xBatch Size=8, Input Frames=16, Resolution=112x1122019.04 | 86.43 | — | — | — | — | |
| TSNBackbone=ResNet-50, Frame=162019.08 | 82.3 | 99.2 | — | — | — | |
| TSNBackbone=ResNet-50, Frame=82019.08 | 81 | 99 | — | — | — | |
| TSN_8FBackbone=ResNet-50, Flow=false, #Frame=8, FLOPs x views=33Gx12020.08 | 81 | 99 | — | — | — | |
| Clip4CLIPevaluation_protocol=kNN, K=202026.02 | — | — | 8.5 | 8.3 | 11.3 | |
| DINOv2evaluation_protocol=kNN, K=202026.02 | — | — | 8.5 | 7.8 | 9.7 | |
| I3Devaluation_protocol=kNN, K=202026.02 | — | — | 26.8 | 25.8 | 53.2 | |
| SemanticMomentsbackbone=DINO, evaluation_protocol=kNN, K=202026.02 | — | — | 25.7 | 25 | 50.8 | |
| SemanticMomentsbackbone=VideoMAE, evaluation_protocol=kNN, K=202026.02 | — | — | 26.9 | 26.5 | 55 | |
| SemanticMomentsbackbone=V-JEPA2, evaluation_protocol=kNN, K=202026.02 | — | — | 28.6 | 28.3 | 47.2 | |
| SlowFastevaluation_protocol=kNN, K=202026.02 | — | — | 19.7 | 17.2 | 34.8 | |
| TimeSFormerevaluation_protocol=kNN, K=202026.02 | — | — | 16.7 | 10.3 | 20.1 | |
| V-JEPA2evaluation_protocol=kNN, K=202026.02 | — | — | 12.5 | 12.3 | 20.7 | |
| VideoMAEevaluation_protocol=kNN, K=202026.02 | — | — | 23.8 | 22.7 | 43.4 | |
| VideoMoCoevaluation_protocol=kNN, K=202026.02 | — | — | 12.1 | 12 | 24.4 | |
| X-CLIPevaluation_protocol=kNN, K=202026.02 | — | — | 26 | 20.2 | 42.2 |