基于注意力增强的跨模态多层融合网络的物体位姿估计

杨恒,王韶涵,董青,赵科渊,杨明亮

Object pose estimation based on attention enhancement cross-modal multilayer fusion network

Heng YANG,Shaohan WANG,Qing DONG,Keyuan ZHAO,Mingliang YANG

表 3 YCB-Video数据集上的效率与精度消融实验

Tab.3 Ablation study on efficiency and accuracy on YCB-Video dataset

模型配置 Np/106 FLOPs/109 tin/ms AUC/% ADD(-S) 分数/% Baseline 78.0 35.9 16.4 83.5 90.1 +Multi-scale 91.0 46.2 21.5 85.8 92.5 +AFE(Attention) 81.0 37.5 17.6 88.9 95.3 完整模型(本文方法) 94.0 47.9 22.3 93.7 97.1