Recurrent MVSNet for High-resolution Multi-view Stereo Depth Inference
This addresses a critical bottleneck for researchers and practitioners in 3D reconstruction by making learned MVS methods scalable to high-resolution scenes, though it is an incremental improvement over existing methods.
The paper tackles the memory scalability limitation in learned multi-view stereo (MVS) approaches by proposing R-MVSNet, which uses a recurrent neural network to sequentially regularize 2D cost maps, reducing memory consumption and enabling high-resolution reconstruction, achieving state-of-the-art performance on MVS benchmarks.
Deep learning has recently demonstrated its excellent performance for multi-view stereo (MVS). However, one major limitation of current learned MVS approaches is the scalability: the memory-consuming cost volume regularization makes the learned MVS hard to be applied to high-resolution scenes. In this paper, we introduce a scalable multi-view stereo framework based on the recurrent neural network. Instead of regularizing the entire 3D cost volume in one go, the proposed Recurrent Multi-view Stereo Network (R-MVSNet) sequentially regularizes the 2D cost maps along the depth direction via the gated recurrent unit (GRU). This reduces dramatically the memory consumption and makes high-resolution reconstruction feasible. We first show the state-of-the-art performance achieved by the proposed R-MVSNet on the recent MVS benchmarks. Then, we further demonstrate the scalability of the proposed method on several large-scale scenarios, where previous learned approaches often fail due to the memory constraint. Code is available at https://github.com/YoYo000/MVSNet.