Unsupervised Learning of Monocular Depth and Ego-Motion Using Multiple Masks

TL;DR

This paper introduces an unsupervised learning approach for monocular depth and ego-motion estimation that uses multiple masks to address occlusion and projection issues, improving accuracy and generalization.

Contribution

It presents a novel masking technique to handle occlusion and projection problems, enhancing the efficiency and accuracy of unsupervised depth and ego-motion learning from monocular videos.

Findings

Achieves good depth and ego-motion performance on KITTI dataset

Demonstrates strong generalization on uncalibrated bike videos

Uses geometric filtering to improve training accuracy

Abstract

A new unsupervised learning method of depth and ego-motion using multiple masks from monocular video is proposed in this paper. The depth estimation network and the ego-motion estimation network are trained according to the constraints of depth and ego-motion without truth values. The main contribution of our method is to carefully consider the occlusion of the pixels generated when the adjacent frames are projected to each other, and the blank problem generated in the projection target imaging plane. Two fine masks are designed to solve most of the image pixel mismatch caused by the movement of the camera. In addition, some relatively rare circumstances are considered, and repeated masking is proposed. To some extent, the method is to use a geometric relationship to filter the mismatched pixels for training, making unsupervised learning more efficient and accurate. The experiments on KITTI dataset show our method achieves good performance in terms of depth and ego-motion. The generalization capability of our method is demonstrated by training on the low-quality uncalibrated bike video dataset and evaluating on KITTI dataset, and the results are still good.