🧭 About Me

Hello! I am Rui Qian, a Master’s student in Computer Control and Automation at Nanyang Technological University (NTU), Singapore, supervised by Prof. Lihua Xie.

My research interests lie at the intersection of 3D Computer Vision and Robotic Perception. My core academic ambition is to advance machine spatial intelligence, focusing on how autonomous agents can perceive, comprehend, and interact with our complex and dynamic 3D world in a manner akin to humans.

I am particularly interested in semantic scene understanding and developing advanced scalable 3D representation methods, such as the involving 3D Gaussian representations. I believe that true autonomy hinges on a machine’s ability to reconstruct (geometry) and understand (semantics) a persistent internal model of its environment.

In the long term, I aim to bridge geometric perception and semantic reasoning. I am passionate about developing foundation models for 3D world understanding to empower future robotic and Embodied AI systems. My goal is to enable these agents with higher-level reasoning capabilities, allowing them to build persistent, high-fidelity digital twins of their environment and ultimately, to seamlessly connect the physical and virtual worlds.

🔥 News

• 2026.02:   🎉🎉 One paper on Embodied Occupancy Networks is accepted to CVPR 2026 !
• 2025.11:   🎉🎉 One paper on Occupancy Networks is accepted to AAAI 2026 as Oral!
• 2025.09:   Currently seeking a PhD position !

📝 Publications

CVPR 2026

TGSFormer: Scalable Temporal Gaussian Splatting for Embodied Semantic Scene Completion

CVPR 2026

Rui Qian^*, Haozhi Cao^*, Tianchen Deng^*, Tianxin Hu, Weixing Guo, Shenghai Yuan, Lihua Xie

We propose TGSFormer, a scalable Temporal Gaussian Splatting framework that maintains persistent Gaussian memory for temporal prediction with dual temporal encoding and confidence-aware voxel fusion. It achieves state-of-the-art results on embodied SSC benchmarks using significantly fewer primitives.

arXiv GitHub Video

AAAI 2026

SplatSSC: Decoupled Depth-Guided Gaussian Splatting for 3D Semantic Scene Completion

AAAI (Oral) 2026

Rui Qian^*, Haozhi Cao^*, Tianchen Deng, Shenghai Yuan, Lihua Xie

We introduce SplatSSC, a decoupled depth-guided Gaussian splatting framework that leverages depth-aware initialization and geometry–semantic disentangled aggregation to achieve highly efficient, robust, and accurate 3D semantic scene completion, outperforming previous methods by over 6.3% IoU and 4.1% mIoU while greatly reducing computational cost.

arXiv GitHub Video

🎖 Honors and Awards

📖 Educations

• 2020.010 - 2024.06, Changzhou University, Changzhou, China, Bachelor of Electronic Information Engineering.
• 2024.08 - now, Nanyang Technological University, Singapore, Master of Computer Control and Automation.

🗺️ Visitor Map