Summer research internship project
Between June and September 2023, I undertook a summer research internship at the University of Southampton in the Vision, Learning, and Control research group, supervised by Dr. Hansung Kim. During this period, my primary focus was on developing a novel system for Real-Time 3D Multi-Person Human Pose Estimation, utilizing an omnidirectional camera and three mmWave radars.
The developed system tackles depth and scale ambiguity issues, operating in real-time with precise performance in both indoor and outdoor environments. This results in an affordable and scalable solution. Notably, my system's time complexity remains nearly constant regardless of the number of detected individuals, achieving a frame rate of approximately 7-8 fps on a laptop with a commercial-grade GPU. It offers simplicity, robustness, and scalability, enhancing technology accessibility and robustness for computer vision applications, and making it an affordable industry solution.
My key contributions include introducing a novel approach for Real-Time 3D Multi-Person Human Pose Estimation, addressing occlusion challenges through an off-the-shelf unsupervised 2D-3D lifting algorithm, conducting camera and radar calibrations, and enhancing the matching of individuals across image and radar spaces. While inspired by preliminary work within our research group, my detection system surpassed its performance.
The culmination of my internship includes the acceptance and presentation of a short paper at the CVMP 2023 Conference and the acceptance of a paper at the ICEIC 2024 Conference. This marks my first ever peer-reviewed publication, where, which is even more impressive, I serve as the first author.