OpenPointCloud

Abstract

This paper gives an overview of OpenPointCloud, the first open-source algorithm library containing outstanding deep learning methods on point cloud compression (PCC). We provide an introduction of our implementations, including 8 methods on lossless geometry PCC and lossy geometry PCC. Principles and contributions of these methods in our algorithm library are illustrated, which are also implemented with different deep learning programming frameworks, such as TensorFlow, Pytorch and TensorLayer. In order to systematically evaluate the performances of all these methods, we conduct a comprehensive benchmarking test. We provide analyses and comparisons of their performances according to their categories and draw constructive conclusions. This algorithm library has been released at https://git.openi.org.cn/OpenPointCloud.