Yijun Yuan (元祎君)

I’m now a PostDoc in MARS lab, led by Prof. Dr. Hang Zhao, in Tsinghua IIIS. I received my PhD (Dr. rer. nat.) from Würzburg University. My dear PhD supervisers are Prof. Dr. Andreas Nüchter, Prof. Dr. Radu Timofte and Prof. Dr. Sören Schwertfeger. Before that I received my Bachelor’s and Master’s Degree from dear ShanghaiTech University. I’m interested in Robotics. My current researches are mainly related to dense SLAM.

CV | ResearchGate | codes

News

Recent researches

SLAM-Former: Putting SLAM into One Transformer Yijun Yuan, Zhuoguang Chen, Kenan Li, Weibang Wang, Hang Zhao paper | website | code We present SLAM-Former, a novel neural approach that integrates full SLAM capabilities into a single transformer.

SceneFactory: A Workflow-centric and Unified Framework for Incremental Scene Modeling Yijun Yuan, Michael Bleier, Andreas Nuchter IEEE Transactions on Robotics (T-RO), 2025 paper | website | code Following the structure of a “Factory”, we introduce workflow-centric framework that provides “assembly lines” for a wide range of applications, to achieve high flexibility, adaptability and production diversification.

Uni-Fusion: Universal Continuous Mapping Yijun Yuan, Andreas Nuchter IEEE Transactions on Robotics (T-RO), 2024 website | paper | code The first universal continuous mapping framework for surfaces, surface properties (color, infrared, style, saliency, etc.) and more (latent features in CLIP embedding space, etc.).

Online Learning of Neural Surface Light Fields alongside Real-time Incremental 3D Reconstruction Yijun Yuan, Andreas Nuchter IEEE Robotics and Automation Letters (RAL), 2023 website | paper | code NSLF-OL explores more fully decoupling the representation of geometry and color. Working alongside real-time surface reconstruction, NSLF-OL focuses on surface field to provide efficient online modeling of light.

An Algorithm for the SE(3)-Transformation on Neural Implicit Maps for Remapping Functions Yijun Yuan, Andreas Nuchter IEEE Robotics and Automation Letters (RAL), 2022 website | paper | code SE(3)-transformation on neural-implicit maps and enable remapping function. This is the first algorithm that make neural-implicit based reconstruction compatible with Loop-closure.

Indirect Point Cloud Registration: Aligning Distance Fields using a Pseudo Third Point Set Yijun Yuan, Andreas Nuchter IEEE Robotics and Automation Letters (RAL), 2022 paper | code Registration two point cloud with a pseudo-third point set. Thus method have potential to work with neural-implicit registration.

Morepage