Learning Depth Estimation for Transparent and Mirror Surfaces

TL;DR

This paper introduces a novel method for depth estimation of transparent and mirror surfaces using pseudo labels generated through in-painting, enabling neural networks to learn without ground-truth annotations.

Contribution

The authors propose a simple, annotation-free pipeline that improves depth estimation for ToM surfaces by generating reliable pseudo labels via in-painting and fine-tuning existing models.

Findings

Significant accuracy improvements on the Booster dataset.

Effective pseudo label generation without ground-truth annotations.

Applicable to both monocular and stereo depth estimation models.

Abstract

Inferring the depth of transparent or mirror (ToM) surfaces represents a hard challenge for either sensors, algorithms, or deep networks. We propose a simple pipeline for learning to estimate depth properly for such surfaces with neural networks, without requiring any ground-truth annotation. We unveil how to obtain reliable pseudo labels by in-painting ToM objects in images and processing them with a monocular depth estimation model. These labels can be used to fine-tune existing monocular or stereo networks, to let them learn how to deal with ToM surfaces. Experimental results on the Booster dataset show the dramatic improvements enabled by our remarkably simple proposal.