Global Contextual Complementary Network for Multi-View Stereo
Yongrong Cao (Ningxia University), Suping Wu (Ningxia University), Xing Zheng (Ningxia University), Bin Wang (Ningxia University),* Pan Li (NingXia University), Zhixiang Yuan (Ningxia University), Lei Lin (Ningxia universiry), Yuxin Peng (Ningxia University)
The 33rd British Machine Vision Conference
Abstract
Multi-View Stereo (MVS) has always been a challenging problem. Existing reconstruction methods mostly rely on convolutional neural networks, which limits the ability of the network to capture the global context of images, resulting in a lack of a certain complete representation of the final depth map. In this paper, we propose a Global Context Complementary Network (GCCN), which aims to enhance the complete representation of depth maps with a global context complementary learning strategy. Specifically, for the feature maps, we first exploit the advantages of convolution neural network (CNN) and self-attention to extract 2D local features and long-term dependence information, respectively. Thus, GCCN achieves maximizing the preservation of the complementary information. Furthermore, in the 3D cost volume regression stage, in order to obtain richer 3D depth information, we design a Contextual-feature Complementary Learning Module (CCLM), which utilizes global feature interaction in the cost volume to achieve complementary learning of cost volumes at different scales. We conduct experiments on the DTU benchmark dataset and the Tanks and Temples dataset. The results show that our approach achieves significant performance compared to state-of-the-art methods.Video
Citation
@inproceedings{Cao_2022_BMVC,
author = {Yongrong Cao and Suping Wu and Xing Zheng and Bin Wang and Pan Li and Zhixiang Yuan and Lei Lin and Yuxin Peng},
title = {Global Contextual Complementary Network for Multi-View Stereo},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0919.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by