Lei Tan, Daole Wang, Miao Yin, Mingyi Sun, Xiuyang Zhao, Xiangyu Kong
{"title":"点云补全的图卷积网络","authors":"Lei Tan, Daole Wang, Miao Yin, Mingyi Sun, Xiuyang Zhao, Xiangyu Kong","doi":"10.1145/3561613.3561616","DOIUrl":null,"url":null,"abstract":"In this paper, we propose a graph convolutional network for point cloud completion (GCNC), predicting the missing region. The local information is embedded into the global features by a graph neural network module, which comprehensively captures relations among points through a graph-based context aggregation. On the other hand, the skip connection effectively utilizes the local structural details of incomplete point clouds during the inference of multi-scale missing parts, which preserves the detailed structure of the complete point cloud. Extensive experiments on the ShapeNet and ModelNet40 benchmark show that the proposed approach outperforms the previous baselines, highlighting its effectiveness. Our quantitative and qualitative analysis confirms the effectiveness of each part of the method.","PeriodicalId":348024,"journal":{"name":"Proceedings of the 5th International Conference on Control and Computer Vision","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"A Graph Convolutional Network for Point Cloud Completion\",\"authors\":\"Lei Tan, Daole Wang, Miao Yin, Mingyi Sun, Xiuyang Zhao, Xiangyu Kong\",\"doi\":\"10.1145/3561613.3561616\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In this paper, we propose a graph convolutional network for point cloud completion (GCNC), predicting the missing region. The local information is embedded into the global features by a graph neural network module, which comprehensively captures relations among points through a graph-based context aggregation. On the other hand, the skip connection effectively utilizes the local structural details of incomplete point clouds during the inference of multi-scale missing parts, which preserves the detailed structure of the complete point cloud. Extensive experiments on the ShapeNet and ModelNet40 benchmark show that the proposed approach outperforms the previous baselines, highlighting its effectiveness. Our quantitative and qualitative analysis confirms the effectiveness of each part of the method.\",\"PeriodicalId\":348024,\"journal\":{\"name\":\"Proceedings of the 5th International Conference on Control and Computer Vision\",\"volume\":\"10 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-08-19\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of the 5th International Conference on Control and Computer Vision\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1145/3561613.3561616\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 5th International Conference on Control and Computer Vision","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/3561613.3561616","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
A Graph Convolutional Network for Point Cloud Completion
In this paper, we propose a graph convolutional network for point cloud completion (GCNC), predicting the missing region. The local information is embedded into the global features by a graph neural network module, which comprehensively captures relations among points through a graph-based context aggregation. On the other hand, the skip connection effectively utilizes the local structural details of incomplete point clouds during the inference of multi-scale missing parts, which preserves the detailed structure of the complete point cloud. Extensive experiments on the ShapeNet and ModelNet40 benchmark show that the proposed approach outperforms the previous baselines, highlighting its effectiveness. Our quantitative and qualitative analysis confirms the effectiveness of each part of the method.