Overview
StreamSplat is a web-based neural-graphics system for streaming and rendering reconstructed scenes in an interactive browser environment. It studies how neural scene representations can be delivered through a hybrid client-server architecture while preserving interactive performance.
Core idea
The system uses a unified color-and-depth interface so scene data can be streamed, fused with depth cues, and rendered on the web. Depth-based fusion connects Gaussian splatting and mesh representations, enabling a more practical hybrid rendering workflow for browser-based neural graphics.
Research contribution
- Hybrid client-server architecture for web neural graphics.
- Color-and-depth streaming interface for reconstructed scenes.
- Depth-based fusion between Gaussian splats and mesh geometry.
- Web-first interaction constraints for neural rendering systems.
Publication
StreamSplat: A Hybrid Client-Server Architecture for Neural Graphics using Depth-based Fusion on the Web
The 30th ACM International Conference on 3D Web Technology (Web3D 2025), Siena, Italy.
Authors: Sehyeon Park, Yechan Yang, Myeongseong Kim, Byounghyun Yoo.
Recognition
The paper received the Best Paper Award at Web3D 2025.
Demo
Materials
Includes system overview diagrams, fidelity comparisons, streaming-compression videos, and interactive scene demos.
- Live demo: https://streamsplat.pengpark.com/