Semi-supervised Representation Learning for 3D Point Clouds

The recent development in the fields of autonomous vehicles, robot vision and virtual reality caused a shift in the research focus - more attention is paid to 3D data representation. In this work, we introduce a novel approach for learning representations for 3D point clouds in semi-supervised mode. The main idea of the approach is to combine the benefits of training autoencoders designed for 3D point clouds in unsupervised mode together with the triplet loss utilized for supervised examples. The proposed method was evaluated considering the classification task and using a challenging benchmark dataset for 3D point clouds.