The KITTI Vision Benchmark Suite

Although depth measurement obtained from LiDAR is
usually sparse, it contains valid and real
distance information, i.e., scale-consistent
absolute depth values. Meanwhile, scale-agnostic
counterparts seek to estimate relative depth and
have achieved impressive performance. To leverage
both the inherent characteristics, we thus suggest
to model scale-consistent depth upon unsupervised
scale-agnostic frameworks. Specifically, we
propose the decomposed scale-consistent learning
(DSCL) strategy, which disintegrates the absolute
depth into relative depth prediction and global
scale estimation, contributing to individual
learning benefits. But unfortunately, most
existing unsupervised scale-agnostic frameworks
heavily suffer from depth holes due to the
extremely sparse depth input and weak supervisory
signal. To tackle this issue, we introduce the
global depth guidance (GDG) module, which
attentively propagates dense depth reference into
the sparse target via novel dense-to-sparse
attention.

http://arxiv.org/abs/2211.10994

@inproceedings{yan2023desnet,
title={Desnet: Decomposed Scale-Consistent
Network for Unsupervised Depth Completion},
author={Yan, Zhiqiang and Wang, Kun and Li,
Xiang and Zhang, Zhenyu and Li, Jun and Yang,
Jian},
booktitle={AAAI (oral)},
volume={37},
number={3},
pages={3109--3117},
year={2023}
}