Augmented Equivariant Mesh Networks for Anatomical Segmentation

Name: Augmented Equivariant Mesh Networks for Anatomical Segmentation
Uploaded: 2026-10-03T00:35:47.277474+00:00
Duration: 329 s
Description: EAMS is a lightweight, equivariant framework for anatomical mesh segmentation built on Equivariant Mesh Neural Networks (EMNN). It uses intrinsic mesh features and augmented message passing to handle irregular geometry robustly across different supervision types. The model achieves competitive performance on diverse tasks while maintaining stability under geometric changes like patient pose.

EAMS is a lightweight, equivariant framework for anatomical mesh segmentation built on Equivariant Mesh Neural Networks (EMNN). It uses intrinsic mesh features and augmented message passing to handle irregular geometry robustly across different supervision types. The model achieves competitive performance on diverse tasks while maintaining stability under geometric changes like patient pose.

Open the full episode