I still can't understand your paper very well due to lack of knowledge of topology.

HuXiaoling / TopoLoss

Code for the NeurIPS 2019 paper: Topology-Preserving Deep Image Segmentation

MIT License

146 stars 22 forks source link

I still can't understand your paper very well due to lack of knowledge of topology. #16

Open Yanfeng-Zhou opened 2 years ago

Yanfeng-Zhou commented 2 years ago

I still can't understand your paper very well due to lack of knowledge of topology. Have you ever written some blogs or guides about this paper? I want to understand the paper more clearly.thank you！

etienne87 commented 1 year ago

I also would like to understand better how the topological loss is differentiable. They give hindsight as to why birth/ death of a particular betti number is differentiable w.r.t pixel value but i am not sure what is the analytical derivation for this. I mean it looks like a heavyside (step) function which to me is not really differentiable (=0 or infinity)