Depth Completion via Deep Basis Fitting
This work addresses depth estimation for computer vision applications, offering incremental improvements over existing methods.
The paper tackles image-guided depth completion by inferring depth from an image and sparse measurements, proposing a method that replaces a final convolutional layer with a least squares fitting module and achieves consistent improvements over state-of-the-art baselines with small computational overhead.
In this paper we consider the task of image-guided depth completion where our system must infer the depth at every pixel of an input image based on the image content and a sparse set of depth measurements. We propose a novel approach that builds upon the strengths of modern deep learning techniques and classical optimization algorithms and significantly improves performance. The proposed method replaces the final $1\times 1$ convolutional layer employed in most depth completion networks with a least squares fitting module which computes weights by fitting the implicit depth bases to the given sparse depth measurements. In addition, we show how our proposed method can be naturally extended to a multi-scale formulation for improved self-supervised training. We demonstrate through extensive experiments on various datasets that our approach achieves consistent improvements over state-of-the-art baseline methods with small computational overhead.