PointMapNet: Point Cloud Feature Map Network for 3D Human Action Recognition

3D human action recognition is crucial in broad industrial application scenarios such as robotics, video surveillance, autonomous driving, or intellectual education, etc. In this paper, we present a new point cloud sequence network called PointMapNet for recognition. PointMapNet, two feature maps symmetrical to depth are proposed summarize appearance and motion representations from sequences. Specifically, first convert the frames virtual using static techniques. The frame 1D vector used characterize structural details frame. Then, inspired by map-based on sequences, symmetrically constructed recognize sequence, i.e., Point Cloud Appearance Map (PCAM) Motion (PCMM). To construct PCAM, an MLP-like architecture designed capture spatio-temporal of sequence. PCMM, difference Finally, map descriptors concatenated fed fully connected classifier order evaluate performance approach, extensive experiments conducted. method achieves impressive results three benchmark datasets, namely NTU RGB+D 60 (89.4% cross-subject 96.7% cross-view), UTD-MHAD (91.61%), MSR Action3D (91.91%). experimental outperform existing state-of-the-art classification networks, demonstrating effectiveness our method.