Learning Depth Estimation for Transparent and Mirror Surfaces
This addresses a specific problem in computer vision for applications like robotics or AR, but is incremental as it builds on existing depth estimation methods.
The paper tackles the challenge of depth estimation for transparent and mirror surfaces by proposing a pipeline that uses pseudo labels from in-painted images to fine-tune existing networks, achieving dramatic improvements on the Booster dataset.
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.