NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID: 3030 Topology-Preserving Deep Image Segmentation

The submission proposes a topology-preserving loss function based on the persistence diagram for image segmentation. After considering the rebuttal, the reviewers were unanimous in their opinion that the submission passed the threshold for acceptance at NeurIPS. The techniques are interesting for image segmentation, and complementary to [25]. It is an interesting combination of topology and statistical learning.