Shape Part

Shape part research focuses on understanding and manipulating the constituent components of 3D shapes, aiming to improve efficiency in tasks like assembly, retrieval, and classification. Current efforts leverage graph neural networks (GNNs) and transformers to learn relationships between parts, often employing self-supervised learning to overcome data limitations and enabling tasks such as part grouping, pose estimation, and semantic segmentation from both geometric and linguistic data. These advancements have significant implications for robotics, CAD design, and manufacturing, promising more efficient and automated processes for creating and manipulating complex 3D objects.