3D Human Pose Estimation on SceneEgo monocular
93MPJPEEgoPoseFormer
Evaluation results
| Method | Links | ||
|---|---|---|---|
| EgoPoseFormerBackbone=ResNet-50, visual feature down-sampling stride=4, transformer decoder layers=3, batch size=16, epochs=12, optimizer=AdamW2024.03 | 93 | 74.3 | |
| SceneEgo2024.03 | 118.5 | 92.7 | |
| EgoPW2024.03 | 189.6 | 105.3 | |
| Mo2Cap²2024.03 | 200.3 | 121.2 | |
| xR-egopose2024.03 | 241.3 | 133.9 |