Digging Into Self-Supervised Monocular Depth Estimation
Per-pixel ground-truth depth data is challenging to acquire at scale. To overcome this limitation, self-supervised learning has emerged as a promising alternative for training models to perform monocular depth estimation. In this paper, we propose a set of improvements, which together result in both quantitatively and qualitatively improved depth maps compared to competing self-supervised methods. Research on self-supervised monocular training usually explores increasingly complex architectures, loss functions, and image formation models, all of which have recently helped to close the gap with fully-supervised methods. We show that a surprisingly simple model, and associated design choices, lead to superior predictions. In particular, we propose (i) a minimum reprojection loss, designed to robustly handle occlusions, (ii) a full-resolution multi-scale sampling method that reduces visual artifacts, and (iii) an auto-masking loss to ignore training pixels that violate camera motion assumptions. We demonstrate the effectiveness of each component in isolation, and show high quality, state-of-the-art results on the KITTI benchmark.
PDF AbstractCode
Results from the Paper
Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Benchmark |
---|---|---|---|---|---|---|
Monocular Depth Estimation | KITTI Eigen split unsupervised | Monodepth2 MS | absolute relative error | 0.106 | # 26 | |
Monocular Depth Estimation | KITTI Eigen split unsupervised | Monodepth2 S | absolute relative error | 0.109 | # 29 | |
Monocular Depth Estimation | KITTI Eigen split unsupervised | Monodepth2 M | absolute relative error | 0.115 | # 36 |