All-day Depth Completion

TL;DR

This paper introduces a multi-sensor depth completion method that combines synthetic data, sparse LiDAR points, and images to improve depth estimation across day and night conditions, especially in low-light regions.

Contribution

It presents SpaDe, a synthetic data-based approach for initial depth approximation, and an uncertainty-driven residual learning scheme for refinement, enhancing depth completion under varying illumination.

Findings

Achieved 11.65% average improvement over baselines in all-day scenarios.

Demonstrated 25% enhancement when augmenting existing methods with SpaDe.

Improved depth estimation accuracy in both daytime and nighttime scenes.

Abstract

We propose a method for depth estimation under different illumination conditions, i.e., day and night time. As photometry is uninformative in regions under low-illumination, we tackle the problem through a multi-sensor fusion approach, where we take as input an additional synchronized sparse point cloud (i.e., from a LiDAR) projected onto the image plane as a sparse depth map, along with a camera image. The crux of our method lies in the use of the abundantly available synthetic data to first approximate the 3D scene structure by learning a mapping from sparse to (coarse) dense depth maps along with their predictive uncertainty - we term this, SpaDe. In poorly illuminated regions where photometric intensities do not afford the inference of local shape, the coarse approximation of scene depth serves as a prior; the uncertainty map is then used with the image to guide refinement through an uncertainty-driven residual learning (URL) scheme. The resulting depth completion network leverages complementary strengths from both modalities - depth is sparse but insensitive to illumination and in metric scale, and image is dense but sensitive with scale ambiguity. SpaDe can be used in a plug-and-play fashion, which allows for 25% improvement when augmented onto existing methods to preprocess sparse depth. We demonstrate URL on the nuScenes dataset where we improve over all baselines by an average 11.65% in all-day scenarios, 11.23% when tested specifically for daytime, and 13.12% for nighttime scenes.