Shape Correspondence

Shape correspondence aims to establish point-wise mappings between geometric shapes, a crucial task in various fields like computer vision and medical imaging. Current research heavily focuses on developing unsupervised methods, leveraging deep learning architectures (e.g., graph neural networks, generative models) and algorithms like optimal transport to find correspondences, particularly for non-rigid and partially observed shapes. These advancements are improving accuracy and efficiency in applications such as shape analysis, object tracking, and medical image registration, enabling more robust and reliable analyses across diverse shape datasets.