ABSTRACT
PointNet++ has gained significant acknowledgement for point cloud data processing capabilities. Over time, various network improvements have been developed to enhance its global learning efficiency, thus boosting the correct segmentation rate. However, these improvements have often resulted in a significant increase in complexity, i.e., the model size and the processing speed. Meanwhile, improvements that focus on complexity reduction while preserving accuracy have been relatively scarce, particularly compared to some simpler models like SqueezeSegV2. To overcome this challenge, we embark on the development of a compact version of the PointNet++ model, namely PointGANet, tailored specifically for three-dimensional point cloud semantic segmentation. In PointGANet, we introduce a grouped attention mechanism in an encoder with grouped convolution incorporated with element-wise multiplication to enrich feature extraction capability and emphasise relevant features. In a decoder, we replace unit pointnet modules with mini pointnet modules to save a massive number of trainable parameters. Through rigorous experimentation, we successfully fine-tune the network to obtain a significant reduction in model size while maintaining accuracy, hence resulting in a substantial enhancement in overall performance. Remarkably, relying on the intensive evaluation using the DALES dataset, PointGANet is more lightweight than the original PointNet++ by approximately five times with some noteworthy improvements in mean accuracy by and mean IoU . These innovations open up exciting possibilities for developing point cloud segmentation applications on IoT and resource-constrained devices.
- R. Qi Charles, Hao Su, Mo Kaichun, and Leonidas J. Guibas. 2017. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Honolulu, HI, USA, 77–85.Google ScholarCross Ref
- Jian-Hong Chen, Guo-Han Lin, Chitra Meghala Yelamandala, and Yu-Cheng Fan. 2020. High-Accuracy Mapping Design Based on Multi-view Images and 3D LiDAR Point Clouds. In 2020 IEEE International Conference on Consumer Electronics (ICCE). IEEE, Las Vegas, NV, USA, 1–2.Google ScholarCross Ref
- Yu-Chen Chou, Yen-Po Lin, Yang-Ming Yeh, and Yi-Chang Lu. 2021. 3D-GFE: a Three-Dimensional Geometric-Feature Extractor for Point Cloud Data. In 2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, Tokyo, Japan, 2013–2017.Google Scholar
- Perry Gibson, José Cano, Jack Turner, Elliot J. Crowley, Michael O’Boyle, and Amos Storkey. 2020. Optimizing Grouped Convolutions on Edge Devices. In 2020 IEEE 31st International Conference on Application-specific Systems, Architectures and Processors (ASAP). IEEE, Manchester, UK, 189–196.Google Scholar
- Yulan Guo, Hanyun Wang, Qingyong Hu, Hao Liu, Li Liu, and Mohammed Bennamoun. 2020. Deep Learning for 3D Point Clouds: A Survey. arxiv:1912.12033 [cs.CV]Google Scholar
- Dong He, Furqan Abid, Young-Min Kim, and Jong-Hwan Kim. 2022. SectorGSnet: Sector Learning for Efficient Ground Segmentation of Outdoor LiDAR Point Clouds. IEEE Access 10 (2022), 11938–11946.Google ScholarCross Ref
- Wang Jisen. 2021. A Study on Target Recognition Algorithm Based on 3D Point Cloud and Feature Fusion. In 2021 IEEE 4th International Conference on Automation, Electronics and Electrical Engineering (AUTEEE). IEEE, Shenyang, China, 630–633.Google ScholarCross Ref
- Huan Lei, Naveed Akhtar, and Ajmal Mian. 2020. SegGCN: Efficient 3D Point Cloud Segmentation With Fuzzy Spherical Kernel. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Seattle, WA, USA, 11608–11617.Google ScholarCross Ref
- Yong Li, Xu Li, Zhenxin Zhang, Feng Shuang, Qi Lin, and Jincheng Jiang. 2022. DenseKPNET: Dense Kernel Point Convolutional Neural Networks for Point Cloud Semantic Segmentation. IEEE Transactions on Geoscience and Remote Sensing 60 (2022), 1–13.Google Scholar
- Chun Liu, Doudou Zeng, Akram Akbar, Hangbin Wu, Shoujun Jia, Zeran Xu, and Han Yue. 2022. Context-Aware Network for Semantic Segmentation Toward Large-Scale Point Clouds in Urban Environments. IEEE Transactions on Geoscience and Remote Sensing 60 (2022), 1–15.Google Scholar
- Xiaofeng Ma, Wei Luo, Mingquan Chen, Jiahui Li, Xin Yan, Xia Zhang, and Wei Wei. 2019. A Fast Point Cloud Segmentation Algorithm Based on Region Growth. In 2019 18th International Conference on Optical Communications and Networks (ICOCN). IEEE, Huangshan, China, 1–2.Google ScholarCross Ref
- Charles R. Qi, Li Yi, Hao Su, and Leonidas J. Guibas. 2017. PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space. (2017). arxiv:1706.02413 [cs.CV]Google Scholar
- Shi Qiu, Saeed Anwar, and Nick Barnes. 2021. Semantic Segmentation for Real Point Cloud Scenes via Bilateral Augmentation and Adaptive Fusion. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Nashville, TN, USA, 1757–1767.Google ScholarCross Ref
- Tianyu Ren and Ruicheng Wu. 2020. An Acceleration Algorithm of 3D Point Cloud Registration Based on Iterative Closet Point. In 2020 Asia-Pacific Conference on Image Processing, Electronics and Computers (IPEC). IEEE, Dalian, China, 271–276.Google ScholarCross Ref
- Gujanatti Rudrappa and Nataraj Vijapur. 2020. Cloud Classification using K-Means Clustering and Content based Image Retrieval Technique. In 2020 International Conference on Communication and Signal Processing (ICCSP). IEEE, Chennai, India, 0700–0704.Google ScholarCross Ref
- Hugues Thomas, Charles R. Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, François Goulette, and Leonidas Guibas. 2019. KPConv: Flexible and Deformable Convolution for Point Clouds. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, Seoul, Korea (South), 6410–6419.Google Scholar
- Nina Varney, Vijayan K. Asari, and Quinn Graehling. 2020. DALES: A Large-scale Aerial LiDAR Data Set for Semantic Segmentation. arxiv:2004.11985 [cs.CV]Google Scholar
- Chongrong Wu, Yitai Lin, Yan Guo, Chenglu Wen, Yongfei Shi, and Cheng Wang. 2022. Vehicle Completion in Traffic Scene Using 3D LiDAR Point Cloud Data. In IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium. IEEE, Kuala Lumpur, Malaysia, 7495–7498.Google ScholarCross Ref
- Wenxuan Wu, Zhongang Qi, and Li Fuxin. 2019. PointConv: Deep Convolutional Networks on 3D Point Clouds. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Long Beach, CA, USA, 9613–9622.Google ScholarCross Ref
- Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. 2022. Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, New Orleans, LA, USA, 19291–19300.Google ScholarCross Ref
- Henan Yuan, Wei Sun, and Tianyuan Xiang. 2020. Line laser point cloud segmentation based on the combination of RANSAC and region growing. In 2020 39th Chinese Control Conference (CCC). IEEE, Shenyang, China, 6324–6328.Google ScholarCross Ref
- Nicolas Tapia Zapata, Nikos Tsoulias, Kowshik Kumar Saha, and Manuela Zude-Sasse. 2022. Fourier analysis of LiDAR scanned 3D point cloud data for surface reconstruction and fruit size estimation. In 2022 IEEE Workshop on Metrology for Agriculture and Forestry (MetroAgriFor). IEEE, Perugia, Italy, 197–202.Google ScholarCross Ref
- Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip Torr, and Vladlen Koltun. 2021. Point Transformer. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, Montreal, QC, Canada, 16239–16248.Google Scholar
- Yiming Zhao, Xiao Zhang, and Xinming Huang. 2021. A Technical Survey and Evaluation of Traditional Point Cloud Clustering Methods for LiDAR Panoptic Segmentation. arxiv:2108.09522 [cs.CV]Google Scholar
- Yu Zhong, Fan Bai, Yong Liu, Lei Huang, Xing Yuan, YuBing Zhang, and JinHang Zhong. 2021. Point Cloud Splicing Based on 3D-Harris Operator. In 2021 3rd International Symposium on Smart and Healthy Cities (ISHC). IEEE, Toronto, ON, Canada, 61–66.Google ScholarCross Ref
- Bing Zhou and Ran Huang. 2020. Segmentation Algorithm for 3D LiDAR Point Cloud Based on Region Clustering. In 2020 7th International Conference on Information, Cybernetics, and Computational Social Systems (ICCSS). IEEE, Guangzhou, China, 52–57.Google ScholarCross Ref
Index Terms
- PointGANet: A Lightweight 3D Point Cloud Learning Architecture for Semantic Segmentation
Recommendations
Strategic Improvements of SqueezeSegV2 for Road-Scene Semantic Segmentation Using 3D LiDAR Point Cloud
SOICT '23: Proceedings of the 12th International Symposium on Information and Communication TechnologySemantic segmentation of LiDAR point clouds for road-scene analysis in autonomous vehicles and driver assistance systems is a challenging task due to the confusion of categories and the sparse distribution of point clouds, thus leading low performance. ...
Deep 3D point cloud classification and segmentation network based on GateNet
AbstractWith the gradual growth of deep learning in machine vision, efficient extraction of 3D point clouds becomes significant. The raw data of the 3D point cloud are sparse, disordered, and immersed in noise, which makes it difficult to classify and ...
Automatic Segmentation of the Prostate on 3D CT Images by Using Multiple Deep Learning Networks
ICBBE '18: Proceedings of the 2018 5th International Conference on Biomedical and Bioinformatics EngineeringAutomatic segmentation of the prostate on CT images has many applications in prostate cancer diagnosis and therapy. However, prostate segmentation from CT images is a very challenging task due to the low contrast of soft tissue and the large variations ...
Comments