3D Point Cloud Semantic Augmentation: Instance Segmentation of 360° Panoramas by Deep Learning Techniques

Semantic augmentation of 3D point clouds is a challenging problem with numerous real-world applications. While deep learning has revolutionised image segmentation and classification, its impact on cloud an active research field. In this paper, we propose instance using architectures. We show the potential indirect approach 2D images Mask R-CNN (Region-Based Convolution Neural Network). Our method consists four core steps. first project onto panoramic three types projections: spherical, cylindrical, cubic. Next, homogenise resulting to correct artefacts empty pixels be comparable available in common training libraries. These are then used as input neural network, designed for segmentation. Finally, obtained predictions reprojected obtain results. link results context-aware network augment semantics. Several tests were performed different datasets test adequacy generalisation. The developed algorithm uses only attributes X, Y, Z, projection centre (virtual camera) position inputs.