Dec, 2023
D3Former: 通过显著性引导的 Transformer 共同学习可重复的稠密探测器和特征增强描述符
D3Former: Jointly Learning Repeatable Dense Detectors and Feature-enhanced Descriptors via Saliency-guided Transformer
Junjie Gao, Pengfei Wang, Qiujie Dong, Qiong Zeng, Shiqing Xin...
TL;DR通过引入一种名为 D3Former 的基于显著性引导的变换器模型,联合学习可重复的稠密检测器和增强特征的描述符,实现了准确匹配点云问题的关键步骤,实验证明该方法在室内和室外基准测试中始终优于最先进的点云匹配方法。