Lei Wu

Ditch the brittle initial segmentation: ArtPro adaptively merges motion proposals during articulated object reconstruction, sidestepping local minima and significantly boosting accuracy.

By explicitly modeling vascular topology, VasGuideNet achieves anatomically consistent Couinaud liver segmentation, surpassing state-of-the-art methods in Dice score and RVD.

By explicitly enforcing forward disease progression, D-ODR ensures that DR grading models learn biologically consistent feature representations, leading to more reliable clinical assessments.

Smooth activation functions unlock smoothness adaptivity in constant-depth neural networks, achieving minimax-optimal statistical rates without needing depth increases like ReLU networks.