Monocular Depth Estimation on KITTI 2015
0.082Abs RelOurs + DinoV2
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Ours + DinoV2Encoder Backbone=DinoV22026.05 | 0.082 | 0.604 | 4.108 | 0.162 | 92.8 | 96.8 | 98.5 | |
| Ours + MonoViTEncoder Backbone=ViT2026.05 | 0.092 | 0.674 | 4.3 | 0.165 | 92.7 | 96.7 | 98.4 | |
| RA-Depth2026.05 | 0.096 | 0.613 | 4.216 | 0.171 | 90.3 | 96.8 | 98.5 | |
| MonoViTEncoder Backbone=ViT2026.05 | 0.099 | 0.708 | 4.372 | 0.175 | 90 | 96.7 | 98.4 | |
| Guizilini et al. (SOTA)Use of off-the-shelf semantic algorithm=true, Architectural restrictions=none2026.05 | 0.1 | 0.761 | 4.27 | 0.175 | 90.2 | 96.5 | 98.2 | |
| DIFFNet2026.05 | 0.102 | 0.764 | 4.483 | 0.18 | 89.6 | 96.5 | 98.3 | |
| Ours (R50)Encoder Backbone=ResNet-502026.05 | 0.104 | 0.725 | 4.453 | 0.18 | 88.5 | 96.2 | 98.3 | |
| Shu et al.Encoder Backbone=ResNet-50, Reproduced results=true, Input Resolution=256x8322026.05 | 0.108 | 0.792 | 4.633 | 0.184 | 88.3 | 96.1 | 98.3 | |
| HR-Depth2026.05 | 0.109 | 0.792 | 4.632 | 0.185 | 88.4 | 96.2 | 98.3 | |
| Monodepth22026.05 | 0.11 | 0.831 | 4.642 | 0.187 | 88.3 | 96.2 | 98.2 | |
| Lego* + LdistEncoder Backbone=ResNet-50, Reproduced results=true, Input Resolution=256x8322026.05 | 0.11 | 0.805 | 4.606 | 0.188 | 88.2 | 96 | 98.2 | |
| Guizilini et al. (R50)Encoder Backbone=ResNet-50, Use of off-the-shelf semantic algorithm=true2026.05 | 0.113 | 0.831 | 4.663 | 0.189 | 87.8 | 97.1 | 98.3 | |
| Lego*Encoder Backbone=ResNet-50, Reproduced results=true, Input Resolution=256x8322026.05 | 0.115 | 0.855 | 4.789 | 0.195 | 87.6 | 95.7 | 97.9 | |
| Lego (paper)2026.05 | 0.154 | 1.272 | 6.012 | 0.23 | — | — | — |