PatchmatchNet: Learned Multi-View Patchmatch Stereo
This work addresses the problem of efficient high-resolution multi-view stereo reconstruction for resource-limited devices, offering significant speed and memory improvements.
This paper introduces PatchmatchNet, a learned cascade Patchmatch formulation for multi-view stereo that processes higher resolution imagery with high computation speed and low memory. It achieves at least 2.5 times faster processing and uses half the memory compared to state-of-the-art methods.
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.