Multiple View Geometry Transformers for 3D Human Pose Estimation
link
时间：CVPR2024
机构：University of Toronto && Southeast University && Microsoft Research Asia
TL;DR
提出一种基于Transformer端到端3D Human Pose Estimation方法MVGFormer，核心模块是geometry与appearance模块。前者是无参可微的，泛化更友好；后者是可学习的，对提高精度更友好。效果SOTA。
Method
Query
Query
标签：Multiple,Geometry,Pose,2D,Human,Estimation,3D From： https://www.cnblogs.com/fariver/p/18357747