Fitting Skeletal Models via Graph-based Learning
This work addresses the problem of automating skeletal model fitting for shape analysis, which is incremental as it builds on existing machine learning-based methods.
The paper tackles the time-consuming manual parameter tuning required for fitting template-based skeletal models by proposing a graph convolutional network method that generates skeletal representations from dense segmentation masks. The method achieves promising results and fast inference on both synthetic data and real hippocampus segmentations.
Skeletonization is a popular shape analysis technique that models an object's interior as opposed to just its boundary. Fitting template-based skeletal models is a time-consuming process requiring much manual parameter tuning. Recently, machine learning-based methods have shown promise for generating s-reps from object boundaries. In this work, we propose a new skeletonization method which leverages graph convolutional networks to produce skeletal representations (s-reps) from dense segmentation masks. The method is evaluated on both synthetic data and real hippocampus segmentations, achieving promising results and fast inference.