作者: Michael B. Holte , Cuong Tran , Mohan M. Trivedi , Thomas B. Moeslund
关键词:
摘要: This paper presents a review and comparative study of recent multi-view 2D 3D approaches for human action recognition. The are reviewed categorized due to their nature. We report comparison the most promising methods using two publicly available datasets: INRIA Xmas Motion Acquisition Sequences (IXMAS) i3DPost Multi-View Human Action Interaction Dataset. Additionally, we discuss some shortcomings camera setups outline our thoughts on future directions