Spatiotemporal prediction on Human3.6M 256x256
108.4MSESimVPv2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SimVPv2Params=11.3M, FLOPs=74.6G, Architecture=Recurrent-free2026.02 | 108.4 | 1,441 | 0.9834 | |
| MogaNetParams=8.6M, FLOPs=63.6G, Architecture=Recurrent-free2026.02 | 109.1 | 1,446.4 | 0.9834 | |
| PredRNN++Params=39.3M, FLOPs=1033.0G, Architecture=Recurrent-based2026.02 | 110 | 1,452.2 | 0.9832 | |
| PFGNetParams=7.3M, FLOPs=58.3G, Architecture=Recurrent-free2026.02 | 111.3 | 1,392.4 | 0.9838 | |
| VANParams=37.5M, FLOPs=182.0G, Architecture=Recurrent-free2026.02 | 111.4 | 1,454.5 | 0.9831 | |
| MIMParams=47.6M, FLOPs=1051.0G, Architecture=Recurrent-based2026.02 | 112.1 | 1,467.1 | 0.9829 | |
| PredRNNParams=24.6M, FLOPs=704.0G, Architecture=Recurrent-based2026.02 | 113.2 | 1,458.3 | 0.9831 | |
| TAUParams=37.6M, FLOPs=182.0G, Architecture=Recurrent-free2026.02 | 113.3 | 1,390.7 | 0.9839 | |
| ConvNeXtParams=31.4M, FLOPs=157.0G, Architecture=Recurrent-free2026.02 | 113.4 | 1,469.7 | 0.9828 | |
| PredRNNv2Params=24.6M, FLOPs=708.0G, Architecture=Recurrent-based2026.02 | 114.9 | 1,484.7 | 0.9827 | |
| SimVPParams=41.2M, FLOPs=197.0G, Architecture=Recurrent-free2026.02 | 115.8 | 1,511.5 | 0.9822 | |
| ConvMixerParams=3.1M, FLOPs=39.4G, Architecture=Recurrent-free2026.02 | 115.8 | 1,527.4 | 0.9822 | |
| UniformerParams=27.7M, FLOPs=211.0G, Architecture=Recurrent-free2026.02 | 116.3 | 1,497.7 | 0.9824 | |
| HorNetParams=28.1M, FLOPs=143.0G, Architecture=Recurrent-free2026.02 | 118.1 | 1,481.1 | 0.9824 | |
| PoolformerParams=31.2M, FLOPs=156.0G, Architecture=Recurrent-free2026.02 | 118.4 | 1,484.1 | 0.9827 | |
| ConvLSTMParams=15.5M, FLOPs=347.0G, Architecture=Recurrent-based2026.02 | 125.5 | 1,566.7 | 0.9813 | |
| PhyDNetParams=4.2M, FLOPs=19.1G, Architecture=Recurrent-based2026.02 | 125.7 | 1,614.7 | 0.9804 | |
| MLP-MixerParams=47.0M, FLOPs=164.0G, Architecture=Recurrent-free2026.02 | 125.7 | 1,511.9 | 0.9819 | |
| MAUParams=20.2M, FLOPs=105.0G, Architecture=Recurrent-based2026.02 | 127.3 | 1,577 | 0.9812 | |
| Swin TransformerParams=38.8M, FLOPs=188.0G, Architecture=Recurrent-free2026.02 | 133.2 | 1,599.7 | 0.9799 | |
| ViTParams=28.3M, FLOPs=239.0G, Architecture=Recurrent-free2026.02 | 136.3 | 1,603.5 | 0.9796 | |
| E3D-LSTMParams=60.9M, FLOPs=542.0G, Architecture=Recurrent-based2026.02 | 143.3 | 1,442.5 | 0.9803 | |
| PredNetParams=12.5M, FLOPs=13.7G, Architecture=Recurrent-based2026.02 | 261.9 | 1,625.3 | 0.9786 |