Xiaodong Gu

A new approach makes regex generation clearer and easier to use.

Investigating differences between human-written and machine-generated code for better detection methods.

VideoMV transforms 2D images and text into fast, consistent 3D visuals.

Hierarchical volume encoding improves 3D shape details and smoothness.

A new method improves surface reconstruction from sparse images, ensuring detail and efficiency.

Explore new methods in generating realistic human movements from text descriptions.

Turn a photo into a moving 3D avatar within minutes.