Research Context

CrossGaussian explores how 3D Gaussian Splatting (3DGS) and real-time 360° streaming can improve remote collaboration.

• 360° video gives remote users a wide view of the local space, but it still lacks depth, free-viewpoint control, and access to occluded areas.
• 3DGS can reconstruct the physical environment as a view-controllable 3D scene, while 360° streaming preserves the live context of the current space.
• The project investigates how these representations can be blended for remote collaboration, including scene blending, occlusion-aware exploration, and cross-reality interaction.

My Contribution

I contributed to the research framing, 3DGS reconstruction workflow, and live 360° sharing prototype for CrossGaussian.

• Researched cross-reality collaboration, 360° telepresence, and 3D reconstruction to frame why panoramic streaming alone is insufficient and why 3DGS is useful for spatial remote collaboration.
• Worked through the 360° capture-to-3DGS reconstruction path: captured physical spaces, prepared footage and image data, ran reconstruction trials with existing 3DGS tooling, and checked whether the reconstructed scenes were usable as remote-collaboration environments.
• Investigated the Insta360 API and Unity integration path, then worked on client-to-client transmission tests for Unity-side 360° video data using TCP/UDP-based approaches so the live camera view could be shared across remote clients.

Research Artifacts

The system overview and Unity implementation captures show how 360° streaming and 3DGS are combined for spatial remote collaboration.

Output

This work contributed to CrossGaussian, a research project published through UIST Adjunct 2025 and ISMAR 2025.

• Paper: CrossGaussian: Enhancing Remote Collaboration through 3D Gaussian Splatting and Real-time 360° Streaming.
• Venue: UIST Adjunct 2025 / ISMAR 2025.