Skeleton Based Action

Skeleton-based action recognition focuses on understanding human actions from sequences of skeletal joint positions, aiming to build robust and efficient systems for various applications. Current research heavily utilizes graph convolutional networks (GCNs) and transformers, often incorporating self-supervised learning techniques like contrastive learning and generative models (e.g., diffusion models) to learn effective representations from often limited labeled data. These advancements are improving the accuracy and efficiency of action recognition, with implications for applications such as human-computer interaction, healthcare (e.g., ADHD diagnosis), and video surveillance.