MMAug, 2018

iSPA-Net: 迭代语义姿态对齐网络

TL;DRiSPA-Net is an iterative Semantic Pose Alignment Network that exploits semantic 3D structural regularity to solve the task of fine-grained pose estimation by predicting viewpoint difference between a given pair of images, and achieves state-of-the-art performance on various real image viewpoint estimation datasets with the aid of correspondence of learned spatial descriptor of the input image pair and refinement in consecutive iterations utilizing an online rendering setup along with effectiveness of a non-uniform bin classification of pose-difference; the approach also shows effectiveness for active object viewpoint localization and unsupervised part-segmentation transfer using only a single part-annotated 3D template model per object class.