ATOM: Self-supervised human action recognition using atomic motion representation learning

Bruno Degardin, Vasco Lopes, Hugo Proença

Abstract: Highlights•Self-supervised atomic motion representation learning scheme for action recognition.•Atomic order prediction as a self-supervised task for skeleton action recognition.•A scalable framework easily integrated into any existing skeleton-based model.•We advance state-of-the-art models in two well-known datasets, NTU-RGBD, NTU120 RGBD.