M^3VSNet: Unsupervised Multi-metric Multi-view Stereo Network

TL;DR

M^3VSNet is an unsupervised multi-metric multi-view stereo network that reconstructs dense 3D point clouds without ground-truth data, using novel loss functions and consistency constraints to achieve competitive results.

Contribution

It introduces a multi-metric loss and normal-depth consistency for unsupervised dense point cloud reconstruction, surpassing previous methods in robustness and generalization.

Findings

Achieves state-of-the-art results among unsupervised MVS methods.

Performs comparably to supervised MVSNet on DTU dataset.

Demonstrates strong generalization on Tanks and Temples benchmark.

Abstract

The present Multi-view stereo (MVS) methods with supervised learning-based networks have an impressive performance comparing with traditional MVS methods. However, the ground-truth depth maps for training are hard to be obtained and are within limited kinds of scenarios. In this paper, we propose a novel unsupervised multi-metric MVS network, named M^3VSNet, for dense point cloud reconstruction without any supervision. To improve the robustness and completeness of point cloud reconstruction, we propose a novel multi-metric loss function that combines pixel-wise and feature-wise loss function to learn the inherent constraints from different perspectives of matching correspondences. Besides, we also incorporate the normal-depth consistency in the 3D point cloud format to improve the accuracy and continuity of the estimated depth maps. Experimental results show that M3VSNet establishes the state-of-the-arts unsupervised method and achieves comparable performance with previous supervised MVSNet on the DTU dataset and demonstrates the powerful generalization ability on the Tanks and Temples benchmark with effective improvement. Our code is available at https://github.com/whubaichuan/M3VSNet