Action Recognition on Diving48 V2
88.1Top-1 AccOG-ReG-B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| OG-ReG-BPretrain=ImageNet-21K+K400, Views=3 × 1, FLOPs=245, Param=96.62026.04 | 88.1 | 98.9 | |
| GC-TDNBackbone=ResNet50, #Frame=162022.03 | 87.6 | — | |
| GC-TSMBackbone=ResNet50, #Frame=162022.03 | 87.2 | — | |
| OG-ReG-BPretrain=ImageNet-21K, Views=3 × 1, FLOPs=245, Param=96.62026.04 | 87 | 98.5 | |
| GC-TSNBackbone=ResNet50, #Frame=162022.03 | 86.8 | — | |
| PST-B†Pretrain=K400, Crops x Clips=3 x 1, FLOPs=252G2022.07 | 86 | 98.6 | |
| DVT-BPretrain=ImageNet-1K, FLOPs=1282026.04 | 86 | — | |
| VIMPACBackbone=Transformer, #Frame=322022.03 | 85.5 | — | |
| PST-BPretrain=K400, Crops x Clips=3 x 1, FLOPs=247G, Params=88.1M2022.07 | 85 | 98.6 | |
| PST-BPretrain=ImageNet-21K+K400, Views=3 × 1, FLOPs=247, Param=88.12026.04 | 85 | 98.6 | |
| OG-ReG-TPretrain=ImageNet-1K+K400, Views=3 × 1, FLOPs=139, Param=61.92026.04 | 84.9 | 98.7 | |
| TDNBackbone=ResNet50, #Frame=16, implementation=our impl.2022.03 | 84.6 | — | |
| OG-ReG-TPretrain=ImageNet-1K, Views=3 × 1, FLOPs=73, Param=32.42026.04 | 83.9 | 98.1 | |
| PST-BPretrain=IN-21K, Crops x Clips=3 x 1, FLOPs=247G2022.07 | 83.6 | 98.5 | |
| PST-BPretrain=ImageNet-21K, Views=3 × 1, FLOPs=247, Param=88.12026.04 | 83.6 | 98.5 | |
| TSMBackbone=ResNet50, #Frame=16, implementation=our impl.2022.03 | 83.2 | — | |
| GC-GSTBackbone=ResNet50, #Frame=162022.03 | 82.5 | — | |
| PST-T†Pretrain=K400, Crops x Clips=3 x 1, FLOPs=74G2022.07 | 82.1 | 98.6 | |
| PST-TPretrain=K400, Crops x Clips=3 x 1, FLOPs=72G, Params=28.5M2022.07 | 81.2 | 98.7 | |
| PST-TPretrain=ImageNet-1K+K400, Views=3 × 1, FLOPs=72, Param=28.52026.04 | 81.2 | 98.7 | |
| TimeSformer-LBackbone=Transformer, #Frame=962022.03 | 81 | — | |
| Timesformer-LPretrain=IN-21K, Crops x Clips=3 x 1, FLOPs=2380G, Params=121.4M2022.07 | 81 | — | |
| Timesformer-LPretrain=ImageNet-21K, Views=3 × 1, FLOPs=2380, Param=121.42026.04 | 81 | — | |
| PST-TPretrain=IN-1K, Crops x Clips=3 x 1, FLOPs=72G2022.07 | 79.2 | 98.2 | |
| PST-TPretrain=ImageNet-1K, Views=3 × 1, FLOPs=72, Param=28.52026.04 | 79.2 | 98.2 | |
| TSNBackbone=ResNet50, #Frame=16, implementation=our impl.2022.03 | 79 | — | |
| GSTBackbone=ResNet50, #Frame=16, implementation=our impl.2022.03 | 78.9 | — | |
| TimeSformer-HRBackbone=Transformer, #Frame=162022.03 | 78 | — | |
| Timesformer-HRPretrain=IN-21K, Crops x Clips=3 x 1, FLOPs=1703G, Params=121.4M2022.07 | 78 | — | |
| Timesformer-HRPretrain=ImageNet-21K, Views=3 × 1, FLOPs=1703, Param=121.42026.04 | 78 | — | |
| SlowFastBackbone=ResNet101, #Frame=64+16, Reference=from [2]2022.03 | 77.6 | — | |
| SlowFast R101, 8x8Pretrain=K400, Crops x Clips=3 x 1, FLOPs=106G, Params=53.3M2022.07 | 77.6 | — | |
| SlowFast-R101, 8x8Pretrain=K400, Views=3 × 1, FLOPs=106, Param=53.32026.04 | 77.6 | — | |
| TimesformerPretrain=IN-21K, Crops x Clips=3 x 1, FLOPs=196G, Params=121.4M2022.07 | 74.9 | — | |
| TimesformerPretrain=ImageNet-21K, Views=3 × 1, FLOPs=196, Param=121.42026.04 | 74.9 | — | |
| FAMEPretrain Dataset=Kinetics-400, Backbone=I3D, Input Resolution=16x224x224, Evaluation Protocol=Finetune2021.09 | 72.9 | — | |
| FAMEPretrain Dataset=UCF101, Backbone=I3D, Input Resolution=16x224x224, Evaluation Protocol=Finetune2021.09 | 67.8 | — | |
| BEPretrain Dataset=Kinetics-400, Backbone=I3D, Input Resolution=16x224x224, Evaluation Protocol=Finetune2021.09 | 62.4 | — | |
| BEPretrain Dataset=UCF101, Backbone=I3D, Input Resolution=16x224x224, Evaluation Protocol=Finetune2021.09 | 58.8 | — | |
| Random Init.Pretrain Dataset=None, Backbone=I3D, Input Resolution=16x224x224, Evaluation Protocol=Finetune2021.09 | 57.4 | — |