Sangyun Lee

Papers

Total Citations

H-Index

About

Sangyun Lee is a researcher whose work focuses on advancing action recognition for robotic perception, a critical challenge in enabling machines to understand and interact with dynamic environments. His key contributions center on improving temporal modeling in video analysis, particularly through the development of the Discriminative Temporal Shift Module (D-TSM), introduced in his 2023 paper. This work addresses the limitations of standard Temporal Shift Modules (TSM) in 2D CNNs, which struggle with complex temporal dynamics in action sequences. By refining how temporal information is shifted and integrated across frames, Lee’s D-TSM enhances the discriminative power of models, making them more effective for real-world robot applications. While his most-cited paper currently holds 4 citations, its impact is growing as the field seeks efficient yet accurate architectures for video understanding. Lee’s research bridges the gap between computational efficiency and performance, offering a practical solution for tasks like human-robot interaction and activity recognition. His work is particularly notable for its focus on lightweight, real-time processing, which is essential for deploying action recognition in resource-constrained robotic systems.