Depth Prediction on OctoSense night (test)
5.99RMSE (m)Late-fusion MAE (w/o RGB)
Evaluation results
| Method | Links | |
|---|---|---|
| Late-fusion MAE (w/o RGB)Removed Modality=RGB2026.06 | 5.99 | |
| Late-fusion MAE (w/o IMU)Removed Modality=IMU2026.06 | 6 | |
| Late-fusion MAE (w/ full attention)Attention=full2026.06 | 6.07 | |
| Late-fusion MAEFusion=late2026.06 | 6.08 | |
| Early-fusion MAEFusion=early2026.06 | 6.3 | |
| Late-fusion MAE (w/o EV)Removed Modality=EV2026.06 | 6.32 | |
| V-JEPA 2.1Encoder=V-JEPA 2.12026.06 | 9.51 | |
| V-JEPA 2.1 (stereo)*Encoder=V-JEPA 2.1, Input=stereo2026.06 | 9.51 | |
| Late-fusion MAE (w/o LiDAR)Removed Modality=LiDAR2026.06 | 9.55 | |
| DINOv2Encoder=DINOv22026.06 | 9.95 | |
| DINOv3Encoder=DINOv32026.06 | 10.03 | |
| Late-fusion MAE (w/ only RGB)Modality=only RGB2026.06 | 10.09 | |
| Perception Enc.Encoder=Perception Enc.2026.06 | 10.2 | |
| RGB-only video MAEModality=RGB, Type=video2026.06 | 10.2 | |
| SigLIP 2Encoder=SigLIP 22026.06 | 10.36 | |
| RGB-only image MAEModality=RGB, Type=image2026.06 | 10.58 |