Xueshi Hou
Currently, I am a Senior Researcher at Tencent, working with the Media Lab led by Dr. Shan Liu. My research lies at the intersection of immersive video, 3D representation, and multimedia systems, with a focus on real-time rendering, streaming, and 3D Gaussian Splatting. I am also actively involved in MPEG standardization.
I received my Ph.D. from the University of California, San Diego, advised by Prof. Sujit Dey, where my research focused on multimedia systems and adaptive streaming. I obtained my B.Eng. from Tsinghua University, advised by Prof. Yong Li, where I worked on mobile networking and edge computing.
Previously, I interned at Samsung, where I worked on immersive media systems including 6DoF video and point cloud streaming.
My goal is to bridge research and real-world deployment of immersive media systems, enabling practical and high-quality user experiences.
Senior Researcher
Research overview
I design multimedia systems that connect 3D representations, interactive rendering, and adaptive delivery. My recent work focuses on practical pipelines that turn immersive media research into deployable systems across web, mobile, and cloud environments.
- 3D Gaussian Splatting for reconstruction and interactive editing
- Immersive video, multiview, and 6DoF streaming systems
- Low-latency multimedia delivery with edge/cloud support
- 3D data and video coding standardization
Experience & education
– Present
- Built end-to-end systems for immersive video and 3D applications, including real-time rendering and streaming (multiview / 6DoF).
- Developed 3D Gaussian Splatting pipelines for reconstruction and interactive rendering across platforms (web, mobile, PC).
- Optimized systems for latency, bandwidth, and visual quality; contributed to MPEG video coding standardization.
- Built a 6DoF spectator mode prototype for eSports with real-time user-controlled viewpoint.
- Implemented end-to-end edge–client streaming over UDP for low-latency control and video transmission.
- Built an adaptive bitrate point cloud streaming system over HTTP for real-time interactive WebVR.
- Designed dynamic level-of-detail selection using Octree structures based on network conditions.
– Dec 2020
- Developed predictive view generation for 360°/VR streaming using head-motion-based deep learning models.
- Designed a hybrid multicast/unicast streaming approach for multi-user VR, improving bandwidth efficiency.
– Jun 2015
- Worked on mobile networking, edge computing, and vehicular networks, including modeling mobility impacts on large-scale urban connectivity.