3D Semantic Segmentation with Submanifold Sparse Convolutional Networks
Convolutionalnetworksarethede-factostandardforanalyzingspatio-temporaldatasuchasimages,videos,and3Dshapes.Whilstsomeofthisdataisnaturallydense(e.g.,photos),manyotherdatasourcesareinherentlysparse.Examplesinclude3DpointcloudsthatwereobtainedusingaLiDARscannerorRGB-Dcamera.Standard“den