Lixin Xue

I am a PhD student at ETH Zürich, supervised by Prof. Otmar Hilliges and Prof. Dimitris Tzionas. My research interests mainly lie in computer vision and computer graphics, especially in the intersection of digital humans and neural fields.

Previously, I completed my master's degree at ETH Zürich. During my master, I did my thesis on neural faces in the Digital Humans Group at DisneyResearch|Studio, supervised by Dr. Paulo Gotardo and Dr. Derek Bradley. Prior to ETH, I did my bachelor's at Beihang University. I was lucky to work with Prof. Hao Su at UCSD and Dr. Zhirong Wu at Microsoft Research Asia.

Email  /  Google Scholar  /  CV  /  Github  /  LinkedIn  /  X

GSTAR enables photo-realistic rendering, surface reconstruction, and 3D tracking for dynamic scenes while handling topology changes.

ODHSR reconstructs a photorealistic dense Gaussian representation of the scene and the moving human within a SLAM setting.

HSR jointly reconstructs dynamic humans and static scenes from monocular RGB videos.

We introduce a novel hybrid representation and end-to-end trainable network architecture to model fully editable and customizable neural avatars.

Make COLMAP more robust to duplicated structures by filtering out wrong matches based on several state-of-the-art algorithms.

A fixed radius nearest neighbors search implemented on CUDA with a similar interface as pytorch3d.ops.knn_points with more than an order of magnitude speedup.

Design a direction-based kernel for post-processing the network predictions to improve performance with limited data.

Achieve better localization accuracy on multiple datasets via a feature-metric loss optimized by LM algorithm.

Website template credits to Jon Barron