The project titled "Multi-Scale Spatio-Temporal Attention Graph Convolutional Network Structure for Skeleton-Based Action Recognition," led by Research Assistant Ugur KILIC from the Faculty of Engineering and Architecture at Erzurum Technical University (ETU), with advisory roles from Dr. Gulshah TUMUKLU OZYER from the Department of Computer Engineering at Ataturk University and Assistant Professor. Ozge OZTIMUR KARADAG from the Department of Computer Engineering at Alanya Alaaddin Keykubat University, has been supported under TUBITAK's Rapid Support Program 1002-A.

In his remarks about the project, Research Assistant Ugur KILIC highlighted that the project aims to develop a new approach in the field of action recognition, which has significant potential applications such as human-computer interaction, augmented reality, robotics vision, among others. The primary goal of the project is to enhance the representation of actions by developing a multi-scale spatio-temporal attention graph convolutional network structure for skeleton-based action recognition. KILIC emphasized that current methods often fail to accurately reflect spatial and temporal dependencies between joints, and the proposed network structure aims to address these challenges by dynamically determining the strength of dependencies between joints through attention mechanisms, thereby updating the skeleton graph structure more accurately. This flexibility in graph construction is expected to enable more detailed and comprehensive analysis of action representations at different scales, thereby improving action recognition performance and facilitating the development of more generalizable models.

Corporate communication Cordinator 22, 06, 2024