Abstract: Skeleton-based human action recognition is becoming increasingly important in a variety of fields. Most existing works train a CNN or GCN based backbone to extract spatial-temporal features, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results