Finally! An AI model that understands SHAPE

Point clouds are the most basic data structure in data science. However, describing the shapes of point clouds holistically is not trivial. Finding a method to describe shape could be a powerful new way to study data manifolds. This fascinating machine learning model is able to describe point cloud shape using a unique approach based on "point nets." 00:00 - Motivation. The difficulty of shape characterization and existing approaches 01:09 - The Yi Shi Model at a high level. Our end goal. 01:56 - Inferring the devil's diet. Analogy for how the model learns. 03:39 - Unpacking the The Yi Shi model. Noising point clouds. 05:27 - More model details 06:16 - Theory behind how the model learns shape descriptors 07: 03 - My dataset and results 08:25 - Outro Original paper this study was based on: "Unsupervised Deep Shape Descriptor with Point Distribution Learning" Yi Shi et al 2020 https://openaccess.thecvf.com/content_CVPR_2020/papers/Shi_Unsupervised_Deep_Shape_Descriptor_With_Point_Distribution_Learning_CVPR_2020_paper.pdf Github Repo: https://github.com/M-Lin-DM/Deep-shape-descriptors-2/tree/master