Novel Action Localization Method Leveraging Skeleton-Snippet Contrastive Learning

This research explores a novel approach to action localization using contrastive learning on skeletal data. The multiscale feature fusion strategy likely enhances performance by capturing action-related information at various temporal granularities.

• •Proposes a novel action localization method.
• •Employs skeleton-snippet contrastive learning.
• •Utilizes multiscale feature fusion.