Rethinking 3D cost aggregation in stereo matching

Wanshui Gan,Wenhao Wu,Shifeng Chen,Yuxiang Zhao,Pak Kin Wong

Pattern Recognition Letters（2023）

引用 2|浏览24

暂无评分

摘要

•Rethinking the 3D convolution in stereo matching tasks with a detailed ablation study•A depth shift module is proposed to reduce the computational cost in cost aggregation•HybridNet has a better trade-off between performance, computational complexity, and model size•The semi-supervised loss could effectively alleviate the artifact problem.

查看译文

关键词

Stereo matching,Disparity estimation,Shift operation,3D Convolution

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要