A Topological Loss Function for Deep-Learning Based Image Segmentation Using Persistent Homology.
James R. CloughNicholas ByrneIlkay ÖksüzVeronika A. ZimmerJulia A. SchnabelAndrew P. KingPublished in: IEEE Trans. Pattern Anal. Mach. Intell. (2022)
Keyphrases
- loss function
- persistent homology
- deep learning
- image segmentation
- topological features
- morse theory
- pairwise
- computational geometry
- unsupervised learning
- support vector
- machine learning
- graph cuts
- markov random field
- weakly supervised
- object segmentation
- multiscale
- higher order
- mental models
- link prediction
- segmentation method
- segmentation algorithm
- shape prior
- region growing
- critical points
- active contours
- learning algorithm
- data points