Quo Vadis, Skeleton Action Recognition?

In this paper, we study current and upcoming frontiers across the landscape of skeleton-based human action recognition. To skeleton-action recognition in wild, introduce Skeletics-152, a curated 3-D pose-annotated subset RGB videos sourced from Kinetics-700, large-scale dataset. We extend our to include out-of-context actions by introducing Skeleton-Mimetics, dataset derived recently introduced Mimetics also Metaphorics, with caption-style annotated YouTube popular social game Dumb Charades interpretative dance performances. benchmark state-of-the-art models on NTU-120 provide multi-layered assessment results. The results benchmarking top performers newly datasets reveal challenges domain gap induced wild. Overall, work characterizes strengths limitations existing approaches datasets. Via datasets, enables new for