I am a junior at Basis Independent Silicon Valley. I am passionate about science and enginnering. I have taken advanced Math and Physics courses in school including AP Calculus BC, AP Physics C, Linear Algebra, Multivariable Calculus and more. I enjoyed doing research and have published two research papers in IEEE International Conferences. Besides STEM, I am a passionate fencer and have won multiple national as well as regional medals. I am a certified fencing refree and have worked in various tournaments and events with fencers from the world. I enjoy playing piano and received state honors in CM10. I also enjoy reading and travelling in my spare time. I have been to about 20 countries to explore culture and history.

Email: william-gordon@outlook.com | Cell: 408-603-1884

• Yuke Xing, William Gordon, Qi Yang, Kaifa Yang, Jiarui Wang, Yiling Xu, “3DGS-VBench: A Comprehensive Video Quality Evaluation Benchmark for 3DGS Compression”, to presented in 2025 IEEE International Conference on Visual Communications and Image Processing, December 1-4, 2025. https://arxiv.org/abs/2508.07038

• Dongkai Hang, William Gordon, Mufan Liu, Le Yang, Qi Yang, Yiling Xu, “Real-Time Viewport-Adaptive Rate Control of Dynamic 3D Gaussian Splatting”, submitted to 2026 IEEE International Conference on Acoustics, Speech, and Signal Processing, May 4-8, 2026.

• William Gordon, Yuke Xing, Kaifa Yang, Yiling Xu, “Video Quality Evaluation Benchmark for 3DGS Compression”, invited by Journal of Electrical and Computational Innovations (JECI), ISSN: 3066-1730, to be submitted.

Abstract:

3D Gaussian Splatting (3DGS) has gained great attention due to its high-fidelity quality and real-time rendering speed. However, the huge 3DGS data volume highlights the new requirement of its transmission under limited bandwidth, hindering the practical deployment of 3DGS. To address this issue, we propose a viewportadaptive rate control strategy for 3DGS that jointly accounts for rendering contribution and aliasing robustness. Given a bandwidth budget, the proposed method dynamically evaluates the set of Gaussian primitives visible from the current viewport and removes the less important ones based on their rendering importance to satisfy the bandwidth constraint. To mitigate the aliasing due to viewport shifts, we further incorporate Mip-Splatting filters during training to preserve the rendering fidelity. Experimental results demonstrate that our method maintains high perceptual quality under constrained bandwidth, and both ablation and trace-driven evaluations confirm its ability to achieve stable, adaptive transmission while preserving visual fidelity.