Jungkwon Lee

Title: Jungkwon Lee and Multi-Camera Object Tracking

Introduction

Jungkwon Lee is an inventor based in Seoul, South Korea, associated with Superb AI Co., Ltd. His patent work addresses the challenge of tracking multiple objects across multiple cameras by combining geometric features with a temporal-spatial graph neural network.

Latest Patents

Lee is named on 1 patent: “Method for Tracking Multiple Objects by Using Multiple Cameras Based on Temporal-Spatial Graph Neural Network Utilizing Geometric Features for Each of the Multiple Objects and Object Tracking Device Using the Same.”

The invention describes a method in which an object-tracking device obtains descriptors for objects detected across cameras and calculates similarities between descriptor pairs. A graph neural network uses nodes and spatial edges to merge detections that correspond to the same object, forming a spatial graph. It then uses correspondence information and temporal edges to connect detections across time, producing a temporal-spatial graph that supports tracking multiple objects within a defined space.

Career Highlights

Lee’s listed patent reflects work on applying graph neural networks to computer vision and multi-object tracking. By combining spatial relationships among detections with temporal correspondences, the method is designed to help identify the same objects across camera views and successive time points.

Collaborations

Lee is associated with Superb AI Co., Ltd. His listed coworker is Eun Soo Im.

Conclusion

Jungkwon Lee’s patent focuses on a graph-based approach to multi-camera object tracking, integrating geometric features and temporal-spatial relationships to follow objects across views and time.