{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/point-cloud-gan","title":"Point Cloud GAN","arxiv_id":"1810.05795","date":"2018-10-13","proceeding":null,"authors":["Chun-Liang Li","Manzil Zaheer","Yang Zhang","Barnabas Poczos","Ruslan Salakhutdinov"],"abstract":"Generative Adversarial Networks (GAN) can achieve promising performance on\nlearning complex data distributions on different types of data. In this paper,\nwe first show a straightforward extension of existing GAN algorithm is not\napplicable to point clouds, because the constraint required for discriminators\nis undefined for set data. We propose a two fold modification to GAN algorithm\nfor learning to generate point clouds (PC-GAN). First, we combine ideas from\nhierarchical Bayesian modeling and implicit generative models by learning a\nhierarchical and interpretable sampling process. A key component of our method\nis that we train a posterior inference network for the hidden variables.\nSecond, instead of using only state-of-the-art Wasserstein GAN objective, we\npropose a sandwiching objective, which results in a tighter Wasserstein\ndistance estimate than the commonly used dual form. Thereby, PC-GAN defines a\ngeneric framework that can incorporate many existing GAN algorithms. We\nvalidate our claims on ModelNet40 benchmark dataset. Using the distance between\ngenerated point clouds and true meshes as metric, we find that PC-GAN trained\nby the sandwiching objective achieves better results on test data than the\nexisting methods. Moreover, as a byproduct, PC- GAN learns versatile latent\nrepresentations of point clouds, which can achieve competitive performance with\nother unsupervised learning algorithms on object recognition task. Lastly, we\nalso provide studies on generating unseen classes of objects and transforming\nimage to point cloud, which demonstrates the compelling generalization\ncapability and potentials of PC-GAN.","url_abs":"http://arxiv.org/abs/1810.05795v1","url_pdf":"http://arxiv.org/pdf/1810.05795v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"point-cloud-gan","repo_url":"https://github.com/chunliangli/Point-Cloud-GAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"object-recognition","task_name":"Object Recognition"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.05795","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}