KwangHyun Ryu

Title: KwangHyun Ryu and Sports Video Activity Recognition

Introduction

KwangHyun Ryu is an inventor associated with Soongsil University. His patent work addresses activity recognition in sports videos, combining attention mechanisms and neural-network processing to identify activities from video data.

Latest Patents

Ryu is associated with one patent: “Apparatus for Recognizing Activity in Sports Video Using Cross Granularity Accumulation Module and Method Thereof.” The invention describes an apparatus and method that use a cross granularity accumulation module (CGAM) to recognize activities in sports videos.

The system first processes video frames through a temporal attention module to distinguish their relative importance. It then passes the frames through convolution blocks and spatial attention modules to produce feature values. The CGAM compresses object information from the spatial attention modules into an object representation, which is combined with the first feature value to extract object features. Finally, a recurrent neural network and a fully connected layer classify the activity using estimated probabilities.

Career Highlights

Ryu’s listed patent contribution focuses on applying attention-based feature extraction and recurrent neural networks to sports video analysis. The patent record provided identifies 1 patent associated with him.

Collaborations

The supplied information lists Soowon Lee as Ryu’s coworker and Soongsil University as his company affiliation.

Conclusion

KwangHyun Ryu’s patent describes a layered approach to sports activity recognition, using temporal and spatial attention alongside cross-granularity feature accumulation and neural-network classification.