GeCo: A Differentiable Geometric Consistency Metric for Video Generation
, Junhwa Hur, Charles Herrmann, Fangneng Zhan, Todd Zickler, ,
Preprint, 2025
I'm a Ph.D. student in Computer Science at , advised by . I also worked closely with at Google DeepMind. Previously, I obtained my M.S. in Biomedical Engineering from , advised by and , and B.S. in Computer Science from the at , advised by .
CV
shixuangu@g.harvard.edu
Our group is looking for RAs/interns, please check this form if you are interested.
Contribute 35mm films for Film Imaging Physics project — receive free high-res scans.
I am broadly interested in 3D computer vision and geometric deep learning. My current research focuses on incorporating physics and geometric priors for visual generation.
Previously, I worked on shape analysis and generative modeling in EM imaging ( [
]) and CT imaging (Rib Analysis).
Additionally, I design and build robots for competition events and medical usage.
[Check the robots I built. ]
[12/2025] Invited talk at [3D Worlds] team, Google DeepMind.
[07/2025] One paper on 3D Curvilinear Structure Segmentation is accecpted by IEEE TMI.
[05/2025] RibFrac Challenge is accecpted by IEEE TMI.
[09/2023] RibSeg v2 is accecpted by IEEE TMI.
[06/2023] One paper on neuron growth simulation is accecpted by CMAME.
* indicates equal contribution.