ResSANet: Learning Geometric Information for Point Cloud Processing
Point clouds with rich local geometric information have potentially huge implications in several applications, especially in areas of robotic manipulation and autonomous driving. However, most point cloud processing methods cannot extract enough geometric features from a raw point cloud, which restr...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-05-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/21/9/3227 |
| _version_ | 1797535102639538176 |
|---|---|
| author | Xiaojun Zhu Zheng Zhang Jian Ruan Houde Liu Hanxu Sun |
| author_facet | Xiaojun Zhu Zheng Zhang Jian Ruan Houde Liu Hanxu Sun |
| author_sort | Xiaojun Zhu |
| collection | DOAJ |
| description | Point clouds with rich local geometric information have potentially huge implications in several applications, especially in areas of robotic manipulation and autonomous driving. However, most point cloud processing methods cannot extract enough geometric features from a raw point cloud, which restricts the performance of their downstream tasks such as point cloud classification, shape retrieval and part segmentation. In this paper, the authors propose a new method where a convolution based on geometric primitives is adopted to accurately represent the elusive shape in the form of a point cloud to fully extract hidden geometric features. The key idea of the proposed approach is building a brand-new convolution net named ResSANet on the basis of geometric primitives to learn hierarchical geometry information. Two different modules are devised in our network, Res-SA and ResSA2, to achieve feature fusion at different levels in ResSANet. This work achieves classification accuracy up to 93.2% on the ModelNet40 dataset and the shape retrieval with an effect of 87.4%. The part segmentation experiment also achieves an accuracy of 83.3% (class mIoU) and 85.3% (instance mIoU) on ShapeNet dataset. It is worth mentioning that the number of parameters in this work is just 1.04 M while the network depth is minimal. Experimental results and comparisons with state-of-the-art methods demonstrate that our approach can achieve superior performance. |
| first_indexed | 2024-03-10T11:39:30Z |
| format | Article |
| id | doaj.art-bbfa48fc64b1456d927100863ff8dc87 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-10T11:39:30Z |
| publishDate | 2021-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-bbfa48fc64b1456d927100863ff8dc872023-11-21T18:36:24ZengMDPI AGSensors1424-82202021-05-01219322710.3390/s21093227ResSANet: Learning Geometric Information for Point Cloud ProcessingXiaojun Zhu0Zheng Zhang1Jian Ruan2Houde Liu3Hanxu Sun4School of Automation, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaCenter for Artificial Intelligence and Robotics, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518005, ChinaCenter for Artificial Intelligence and Robotics, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518005, ChinaCenter for Artificial Intelligence and Robotics, Shenzhen International Graduate School, Tsinghua University, Shenzhen 518005, ChinaSchool of Automation, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaPoint clouds with rich local geometric information have potentially huge implications in several applications, especially in areas of robotic manipulation and autonomous driving. However, most point cloud processing methods cannot extract enough geometric features from a raw point cloud, which restricts the performance of their downstream tasks such as point cloud classification, shape retrieval and part segmentation. In this paper, the authors propose a new method where a convolution based on geometric primitives is adopted to accurately represent the elusive shape in the form of a point cloud to fully extract hidden geometric features. The key idea of the proposed approach is building a brand-new convolution net named ResSANet on the basis of geometric primitives to learn hierarchical geometry information. Two different modules are devised in our network, Res-SA and ResSA2, to achieve feature fusion at different levels in ResSANet. This work achieves classification accuracy up to 93.2% on the ModelNet40 dataset and the shape retrieval with an effect of 87.4%. The part segmentation experiment also achieves an accuracy of 83.3% (class mIoU) and 85.3% (instance mIoU) on ShapeNet dataset. It is worth mentioning that the number of parameters in this work is just 1.04 M while the network depth is minimal. Experimental results and comparisons with state-of-the-art methods demonstrate that our approach can achieve superior performance.https://www.mdpi.com/1424-8220/21/9/3227point-cloud processingdeep neural networksmachine learninggeometric primitives |
| spellingShingle | Xiaojun Zhu Zheng Zhang Jian Ruan Houde Liu Hanxu Sun ResSANet: Learning Geometric Information for Point Cloud Processing Sensors point-cloud processing deep neural networks machine learning geometric primitives |
| title | ResSANet: Learning Geometric Information for Point Cloud Processing |
| title_full | ResSANet: Learning Geometric Information for Point Cloud Processing |
| title_fullStr | ResSANet: Learning Geometric Information for Point Cloud Processing |
| title_full_unstemmed | ResSANet: Learning Geometric Information for Point Cloud Processing |
| title_short | ResSANet: Learning Geometric Information for Point Cloud Processing |
| title_sort | ressanet learning geometric information for point cloud processing |
| topic | point-cloud processing deep neural networks machine learning geometric primitives |
| url | https://www.mdpi.com/1424-8220/21/9/3227 |
| work_keys_str_mv | AT xiaojunzhu ressanetlearninggeometricinformationforpointcloudprocessing AT zhengzhang ressanetlearninggeometricinformationforpointcloudprocessing AT jianruan ressanetlearninggeometricinformationforpointcloudprocessing AT houdeliu ressanetlearninggeometricinformationforpointcloudprocessing AT hanxusun ressanetlearninggeometricinformationforpointcloudprocessing |