Featured Stories Grant Stories News

4-VA@Mason Computing Team Gets Ahead of the Game


While some professors worry that they are two steps behind technology, scholars in Mason’s Department of Computer Science pride themselves on staying two steps ahead.  However, that wasn’t enough for Associate Professor Bo Han, who wanted to deepen the educational experience for students.

When Han’s proposal “Innovating Point Cloud Processing for Networked Systems” was approved for pilot funding by 4-VA@Mason, he brought together co-PI Felix Xiaozhu Lin, Associate Professor, Department of Computer Science in the UVA School of Engineering, specialists from the University of Minnesota (UM), the New Jersey Institute of Technology (NJIT), and graduate and undergraduate students taking his new class.

It was an “all-hands-on-deck” project with each of the team members tackling a variety of challenges, resulting in the delivery of a product they dubbed DeepMix, a mobility-aware, lightweight, and hybrid 3D object detection framework. A unique feature of DeepMix is that it fully utilizes the mobility of headsets to fine-tune detection results and boosts detection precision. In fact, when Han’s team implemented a prototype of DeepMix on Microsoft HoloLens and evaluated its performance through both extensive controlled experiments and a user study with more than 30 participants, DeepMix not only improved detection accuracy by 9.1 to 37.3% but significantly boosted detection accuracy in mobile scenarios.

Han credits the success of this project to his collaboration with partners at UVA, UM, NJIT, and especially his students.  These include graduate students Nan Wu, who led the design and implementation of point cloud super-resolution for 3D object detection; along with Ruizhi Cheng and Puqi Zhou, who worked on the implementation and evaluation of gaze-assisted motion prediction for point cloud streaming. Undergraduate student Jing Wang also participated in the project, handling the implementation of the back-end system for image-based localization to improve the accuracy of pose estimation and motion prediction for point cloud streaming.

Pictured: Nan Wu presenting one of the project’s resultant published papers at the Association for Computing Machinery HotMobile 2022 workshop.

“This project was high intensity, multi-faceted, and challenging, but thanks to our 4-VA@Mason grant we were able to develop a great team and produce concrete results,” says Han.  “Now, we want to move our technology to the next level and build interactive holographic communication systems for truly immersive remote collaboration based on mixed reality.