Video Prediction on Cityscapes 128x128 resolution (test)
141.31FVDMCVD concat past-mask
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MCVD concat past-maskk=52022.05 | 141.31 | 0.112 | 69 | |
| MCVD spatin past-maskk=52022.05 | 184.81 | 0.121 | 72 | |
| GHVAEk=102022.05 | 418 | 0.193 | 74 | |
| Hierarchical VRNN (Ours w/ Hier)Resolution=128x128, Number of predicted frames=28, Number of samples per test sequence=1002019.04 | 567.51 | 0.264 | 62.8 | |
| Hier-vRNNk=102022.05 | 567.51 | 0.264 | 62.8 | |
| Improved Conditional VRNN (Ours w/o Hier)Resolution=128x128, Number of predicted frames=28, Number of samples per test sequence=1002019.04 | 682.08 | 0.304 | 60.9 | |
| vRNN ILk=102022.05 | 682.08 | 0.304 | 60.9 | |
| SVG-LPResolution=128x128, Number of predicted frames=28, Number of samples per test sequence=1002019.04 | 1,300.26 | 0.549 | 57.4 | |
| SVG-LPk=102022.05 | 1,300.26 | 0.549 | 57.4 |