SuperDepth: Self-Supervised, Super-Resolved Monocular Depth Estimation

TL;DR

This paper introduces a high-resolution self-supervised monocular depth estimation method called SuperDepth, which leverages super-resolution techniques and flip-augmentation to improve depth prediction accuracy.

Contribution

It proposes a novel depth super-resolution layer and a differentiable flip-augmentation layer, significantly enhancing self-supervised depth and pose estimation performance.

Findings

Achieves state-of-the-art results on KITTI benchmark

Improves depth prediction fidelity at high resolution

Reduces artifacts from occlusions in disparity maps

Abstract

Recent techniques in self-supervised monocular depth estimation are approaching the performance of supervised methods, but operate in low resolution only. We show that high resolution is key towards high-fidelity self-supervised monocular depth prediction. Inspired by recent deep learning methods for Single-Image Super-Resolution, we propose a sub-pixel convolutional layer extension for depth super-resolution that accurately synthesizes high-resolution disparities from their corresponding low-resolution convolutional features. In addition, we introduce a differentiable flip-augmentation layer that accurately fuses predictions from the image and its horizontally flipped version, reducing the effect of left and right shadow regions generated in the disparity map due to occlusions. Both contributions provide significant performance gains over the state-of-the-art in self-supervised depth and pose estimation on the public KITTI benchmark. A video of our approach can be found at https://youtu.be/jKNgBeBMx0I.