RC-MVSNet: Unsupervised Multi-View Stereo with Neural Rendering

TL;DR

RC-MVSNet introduces a neural rendering-based unsupervised multi-view stereo method that effectively handles non-Lambertian surfaces and occlusions, achieving state-of-the-art results without supervision.

Contribution

It proposes a novel neural rendering framework with depth consistency and view synthesis losses for unsupervised MVS, addressing challenges of non-Lambertian surfaces and occlusions.

Findings

Achieves state-of-the-art unsupervised MVS performance on DTU and Tanks&Temples.

Performs competitively with supervised MVS methods.

Effectively handles non-Lambertian surfaces and occlusions.

Abstract

Finding accurate correspondences among different views is the Achilles' heel of unsupervised Multi-View Stereo (MVS). Existing methods are built upon the assumption that corresponding pixels share similar photometric features. However, multi-view images in real scenarios observe non-Lambertian surfaces and experience occlusions. In this work, we propose a novel approach with neural rendering (RC-MVSNet) to solve such ambiguity issues of correspondences among views. Specifically, we impose a depth rendering consistency loss to constrain the geometry features close to the object surface to alleviate occlusions. Concurrently, we introduce a reference view synthesis loss to generate consistent supervision, even for non-Lambertian surfaces. Extensive experiments on DTU and Tanks\&Temples benchmarks demonstrate that our RC-MVSNet approach achieves state-of-the-art performance over unsupervised MVS frameworks and competitive performance to many supervised methods.The code is released at https://github.com/Boese0601/RC-MVSNet