Fourier analysis on robustness of graph convolutional neural networks for skeleton-based action recognition

Nariki Tanaka, Hiroshi Kera, Kazuhiko Kawamoto

Abstract: Highlights•Novel Fourier analysis in GCN robustness for skeletal action recognition.•Discovery of no robustness trade-off in adversarially-trained GCNs.•Adversarially-trained GCNs robust to Gaussian corruption and frame loss.