PatchmatchNet: Learned Multi-View Patchmatch Stereo

TL;DR

PatchmatchNet introduces a learnable, efficient multi-view stereo method that leverages an iterative multi-scale Patchmatch approach, achieving high accuracy with significantly reduced computational resources and memory usage.

Contribution

The paper presents a novel end-to-end trainable architecture that enhances Patchmatch with adaptive propagation, enabling high-resolution multi-view stereo processing on resource-limited devices.

Findings

Achieves at least 2.5x faster processing than state-of-the-art methods.

Uses half the memory of existing top models.

Demonstrates competitive accuracy and generalization across multiple datasets.

Abstract

We present PatchmatchNet, a novel and learnable cascade formulation of Patchmatch for high-resolution multi-view stereo. With high computation speed and low memory requirement, PatchmatchNet can process higher resolution imagery and is more suited to run on resource limited devices than competitors that employ 3D cost volume regularization. For the first time we introduce an iterative multi-scale Patchmatch in an end-to-end trainable architecture and improve the Patchmatch core algorithm with a novel and learned adaptive propagation and evaluation scheme for each iteration. Extensive experiments show a very competitive performance and generalization for our method on DTU, Tanks & Temples and ETH3D, but at a significantly higher efficiency than all existing top-performing models: at least two and a half times faster than state-of-the-art methods with twice less memory usage.