spatially专题

《Neural Pose Transfer by Spatially Adaptive Instance Normalization》

概括: 对identity mesh 做pose transfer, 实际上是让提供目标pose的human mesh 学到identity的style。关键在于要让生成的mesh的点的顺序和identity mesh的一致而不是pose mesh。   传统的做法是学习在同一个pose下,不同identity style 之间的correspondence,利用correspondence

论文阅读:《Fast Convergence of DETR with Spatially Modulated Co-Attention》

论文链接:https://arxiv.org/pdf/2101.07448.pdf. arxiv 2021 文章目录 1 Background2 Motivation3 Related Work4 Advantages/Contributions5 Method5.1. A Revisit of DETR5.2. Spatially Modulated Co-Attention 6 Exp