Shaoxiang Wang (王绍翔)

Shaoxiang Wang (王绍翔)

Shaoxiang Wang

I am a researcher in the Augmented Reality department at German Research Center for AI (DFKI) in Germany, under the supervision of Alain Pagani and Didier Stricker. My research topic is 3D vision and reconstruction. Feel free to reach out if you’re interested in my concurrent research or just want to have a casual chat! :D

I received my Master’s degree in Mechatronics from the Karlsruhe Institute of Technology (KIT), during which I conducted a 6-month thesis project in efficient computer vision at BMW Group with Lukas Frickenstein, under the supervision of Jürgen Becker at KIT and Walter Stechele at TU Munich. Prior to KIT, I obtained a Bachelor’s degree in Mechatronics from Harbin Institute of Technology (HIT).

Publications

ReLaGS: Relational Language Gaussian Splatting

CVPR 2026

[arXiv] [code] [website]

A platform for open-vocabulary 3D reasoning with hierarchical semantics and relational scene graphs.

Inpaint360GS: Efficient Object-Aware 3D Inpainting via Gaussian Splatting for 360° Scenes

Shaoxiang Wang, Shihong Zhang, Christen Millerdurai, Didier Stricker, Alain Pagani

WACV 2026

[arXiv] [code] [website]

We enable multi-object removal and consistent 360° scene completion for your scene!

TalkingPose: Efficient Face and Gesture Animation with Feedback-guided Diffusion Model

Alireza Javanmardi, Tewodros Amberbir Habtegebrial, Pragati Jaiswal, Christen Millerdurai, Shaoxiang Wang, Alain Pagani, Didier Stricker

WACV 2026

[arXiv] [code] [website]

Achieve temporally consistent long-form animation via feedback-driven diffusion.

Uni-SLAM: Uncertainty-Aware Neural Implicit SLAM for Real-Time Dense Indoor Scene Reconstruction

Shaoxiang Wang, Yaxu Xie, Chun-Peng Chang, Christen Millerdurai, Alain Pagani, Didier Stricker

WACV 2025

[arXiv] [code] [website]

Reweight supervision with predictive uncertainty and adapt mapping to achieve accurate real-time indoor reconstruction.

MiKASA: Multi-Key-Anchor & Scene-Aware Transformer for 3D Visual Grounding

Chun-Peng Chang, Shaoxiang Wang, Alain Pagani, Didier Stricker

CVPR 2024

[arXiv] [code] [website]

We introduce multi-key-anchor attention to better capture spatial relationships.