An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression
Due to the tremendous volume taken by the 3D point-cloud models, knowing how to achieve the balance between a high compression ratio, a low distortion rate, and computing cost in point-cloud compression is a significant issue in the field of virtual reality (VR). Convolutional neural networks have b...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-02-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/4/2250 |
| _version_ | 1827755485117284352 |
|---|---|
| author | Guoliang Luo Bingqin He Yanbo Xiong Luqi Wang Hui Wang Zhiliang Zhu Xiangren Shi |
| author_facet | Guoliang Luo Bingqin He Yanbo Xiong Luqi Wang Hui Wang Zhiliang Zhu Xiangren Shi |
| author_sort | Guoliang Luo |
| collection | DOAJ |
| description | Due to the tremendous volume taken by the 3D point-cloud models, knowing how to achieve the balance between a high compression ratio, a low distortion rate, and computing cost in point-cloud compression is a significant issue in the field of virtual reality (VR). Convolutional neural networks have been used in numerous point-cloud compression research approaches during the past few years in an effort to progress the research state. In this work, we have evaluated the effects of different network parameters, including neural network depth, stride, and activation function on point-cloud compression, resulting in an optimized convolutional neural network for compression. We first have analyzed earlier research on point-cloud compression based on convolutional neural networks before designing our own convolutional neural network. Then, we have modified our model parameters using the experimental data to further enhance the effect of point-cloud compression. Based on the experimental results, we have found that the neural network with the 4 layers and 2 strides parameter configuration using the <i>Sigmoid</i> activation function outperforms the default configuration by 208% in terms of the compression-distortion rate. The experimental results show that our findings are effective and universal and make a great contribution to the research of point-cloud compression using convolutional neural networks. |
| first_indexed | 2024-03-11T08:10:00Z |
| format | Article |
| id | doaj.art-8d318a27a0b746f883b48e99fc8351af |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-11T08:10:00Z |
| publishDate | 2023-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-8d318a27a0b746f883b48e99fc8351af2023-11-16T23:12:18ZengMDPI AGSensors1424-82202023-02-01234225010.3390/s23042250An Optimized Convolutional Neural Network for the 3D Point-Cloud CompressionGuoliang Luo0Bingqin He1Yanbo Xiong2Luqi Wang3Hui Wang4Zhiliang Zhu5Xiangren Shi6Virtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaVirtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaVirtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaVirtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaVirtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaVirtual Reality and Interactive Techniques Institute, East China Jiaotong University, Nanchang 330013, ChinaSchool of Informatics, Xiamen University, Xiamen 361005, ChinaDue to the tremendous volume taken by the 3D point-cloud models, knowing how to achieve the balance between a high compression ratio, a low distortion rate, and computing cost in point-cloud compression is a significant issue in the field of virtual reality (VR). Convolutional neural networks have been used in numerous point-cloud compression research approaches during the past few years in an effort to progress the research state. In this work, we have evaluated the effects of different network parameters, including neural network depth, stride, and activation function on point-cloud compression, resulting in an optimized convolutional neural network for compression. We first have analyzed earlier research on point-cloud compression based on convolutional neural networks before designing our own convolutional neural network. Then, we have modified our model parameters using the experimental data to further enhance the effect of point-cloud compression. Based on the experimental results, we have found that the neural network with the 4 layers and 2 strides parameter configuration using the <i>Sigmoid</i> activation function outperforms the default configuration by 208% in terms of the compression-distortion rate. The experimental results show that our findings are effective and universal and make a great contribution to the research of point-cloud compression using convolutional neural networks.https://www.mdpi.com/1424-8220/23/4/2250point-cloud compressionconvolutional neural networkactivation function |
| spellingShingle | Guoliang Luo Bingqin He Yanbo Xiong Luqi Wang Hui Wang Zhiliang Zhu Xiangren Shi An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression Sensors point-cloud compression convolutional neural network activation function |
| title | An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression |
| title_full | An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression |
| title_fullStr | An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression |
| title_full_unstemmed | An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression |
| title_short | An Optimized Convolutional Neural Network for the 3D Point-Cloud Compression |
| title_sort | optimized convolutional neural network for the 3d point cloud compression |
| topic | point-cloud compression convolutional neural network activation function |
| url | https://www.mdpi.com/1424-8220/23/4/2250 |
| work_keys_str_mv | AT guoliangluo anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT bingqinhe anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT yanboxiong anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT luqiwang anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT huiwang anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT zhiliangzhu anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT xiangrenshi anoptimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT guoliangluo optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT bingqinhe optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT yanboxiong optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT luqiwang optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT huiwang optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT zhiliangzhu optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression AT xiangrenshi optimizedconvolutionalneuralnetworkforthe3dpointcloudcompression |