Video Prediction on Moving-MNIST 10 → 10 (test)
15.05MSESimVPv2-S×10
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SimVPv2-S×10FLOPs (G)=16.53, Training time (total s)=1560, Inference efficiency (FPS)=44.09, Training Epochs=20002022.11 | 15.05 | 0.967 | — | 49.8 | |
| SimVPv2-S×5FLOPs (G)=16.53, Training time (total s)=780, Inference efficiency (FPS)=44.09, Training Epochs=10002022.11 | 16.47 | 0.964 | — | 53.24 | |
| Temporal Attention Unit2022.06 | 19.8 | 0.957 | — | 60.3 | |
| SimVPv2-LFLOPs (G)=152.2, Training time (s/epoch)=796, Inference efficiency (FPS)=21.232022.11 | 21.81 | 0.952 | — | 66.43 | |
| Crevnet2022.06 | 22.3 | 0.949 | — | — | |
| SimVPv2-S×3FLOPs (G)=16.53, Training time (total s)=468, Inference efficiency (FPS)=44.09, Training Epochs=6002022.11 | 22.37 | 0.951 | — | 67.52 | |
| PredRNN++FLOPs (G)=171.73, Training time (s/epoch)=1280, Inference efficiency (FPS)=3.712022.11 | 22.45 | 0.95 | — | 69.7 | |
| MIMFLOPs (G)=179.18, Training time (s/epoch)=1388, Inference efficiency (FPS)=3.082022.11 | 23.66 | 0.946 | — | 74.37 | |
| SimVP2022.06 | 23.8 | 0.948 | — | 68.9 | |
| PhyDNet2022.06 | 24.4 | 0.947 | — | 70.3 | |
| PredRNNFLOPs (G)=115.95, Training time (s/epoch)=869, Inference efficiency (FPS)=3.972022.11 | 25.04 | 0.944 | — | 76.26 | |
| SimVPv2-SFLOPs (G)=16.53, Training time (s/epoch)=156, Inference efficiency (FPS)=44.092022.11 | 26.6 | 0.94 | — | 77.32 | |
| SwinLSTMFLOPs (G)=69.87, Training time (s/epoch)=820, Inference efficiency (FPS)=6.512022.11 | 27.44 | 0.938 | — | 78.69 | |
| PredRNNv2FLOPs (G)=116.59, Training time (s/epoch)=899, Inference efficiency (FPS)=3.492022.11 | 27.73 | 0.937 | — | 82.17 | |
| ConvLSTM-LFLOPs (G)=127.01, Training time (s/epoch)=879, Inference efficiency (FPS)=6.242022.11 | 29.88 | 0.925 | — | 95.05 | |
| CrevNetFLOPs (G)=270.68, Training time (s/epoch)=1166, Inference efficiency (FPS)=1.012022.11 | 30.15 | 0.935 | — | 86.28 | |
| MAUFLOPs (G)=17.79, Training time (s/epoch)=535, Inference efficiency (FPS)=3.082022.11 | 30.64 | 0.928 | — | 88.17 | |
| SimVPFLOPs (G)=19.43, Training time (s/epoch)=261, Inference efficiency (FPS)=27.152022.11 | 32.22 | 0.927 | — | 89.19 | |
| MMVPFLOPs (G)=93.55, Training time (s/epoch)=402, Inference efficiency (FPS)=21.332022.11 | 33.29 | 0.926 | — | 89.61 | |
| PhyDNetFLOPs (G)=15.33, Training time (s/epoch)=452, Inference efficiency (FPS)=4.622022.11 | 35.68 | 0.917 | — | 96.7 | |
| E3D-LSTMFLOPs (G)=298.87, Training time (s/epoch)=2693, Inference efficiency (FPS)=3.732022.11 | 36.19 | 0.932 | — | 78.64 | |
| DDPAE2022.06 | 38.9 | 0.922 | — | 90.7 | |
| E3D-LSTM2022.06 | 41.3 | 0.91 | — | 87.2 | |
| LMC-Memory2021.04 | 41.5 | 0.924 | 46.9 | — | |
| LMC2022.06 | 41.5 | 0.924 | — | — | |
| PredRNN++2021.04 | 42.1 | 0.913 | 59.5 | — | |
| MIM2022.06 | 44.2 | 0.91 | — | 101.1 | |
| ConvLSTM-SFLOPs (G)=14.45, Training time (s/epoch)=190, Inference efficiency (FPS)=7.52022.11 | 46.26 | 0.878 | — | 142.18 | |
| PredRNN++2022.06 | 46.5 | 0.898 | — | 106.8 | |
| E3D-LSTM2021.04 | 50.9 | 0.912 | 86.7 | — | |
| Conv-TT-LSTM2021.04 | 53 | 0.915 | 40.5 | — | |
| Conv-TT-LSTM2022.06 | 53 | 0.915 | — | — | |
| PredRNN2021.04 | 56.8 | 0.867 | — | — | |
| PredRNN2022.06 | 56.8 | 0.867 | — | 126.1 | |
| VPN2021.04 | 64.1 | 0.87 | — | — | |
| VPN2022.06 | 64.1 | 0.87 | — | — | |
| CDNA2021.04 | 97.4 | 0.721 | — | — | |
| ConvLSTM2022.06 | 103.3 | 0.707 | — | 182.9 | |
| TRAJGRU2021.04 | 106.9 | 0.713 | — | — |