Email this article Setting up model training for classification and segmentation of Point Clouds
- Authors: Gura D.А.1,2, Dyachenko R.A.1, Boyko E.S.1,3, Levchenko D.A.3
-
Affiliations:
- Kuban State Technological University
- Kuban State Agrarian University
- Kuban State University
- Issue: No 1 (2024)
- Pages: 92-102
- Section: Machine Learning, Neural Networks
- URL: https://journals.rcsi.science/2071-8594/article/view/269788
- DOI: https://doi.org/10.14357/20718594240108
- EDN: https://elibrary.ru/WMZHVG
- ID: 269788
Cite item
Full Text
Abstract
The features and capabilities of the PointNet neural network architecture in relation to artificially generated clouds of laser reflection points in the Terra_Maker information system are presented. The results of training by the Paintnet network are analyzed and the accuracy of the obtained models and graphs is evaluated. An approach is proposed to determine the parameters that give maximum accuracy when performing experiments on the example of point clouds obtained from the Terra_Maker information system.
Full Text
About the authors
Dmitry А. Gura
Kuban State Technological University; Kuban State Agrarian University
Author for correspondence.
Email: gda-kuban@mail.ru
Candidate of Technical Sciences, Assistant professor, Assistant professor
Russian Federation, Krasnodar; KrasnodarRoman A. Dyachenko
Kuban State Technological University
Email: emessage@rambler.ru
Doctor of Technical Sciences, Professor of the Department of Computer Science and Computer Engineering
Russian Federation, KrasnodarEvgeny S. Boyko
Kuban State Technological University; Kuban State University
Email: boykoes@yandex.ru
Candidate of Geographical Sciences Assistant professor of the Department of Geoinformatics
Russian Federation, Krasnodar; KrasnodarDmitry Alexandrovich Levchenko
Kuban State University
Email: levchenkodima@mail.ru
Candidate of Pedagogical Sciences, Assistant professor of the Department of Data Analysis and Artificial Intelligence
Russian Federation, KrasnodarReferences
- Zhu H. H., Tuya D., Mou L., Xia G. S., Zhang L., Xu F., Fraundorfer F. Deep learning in Remote sensing: a comprehensive overview and list of resources // IEEE Geoscience and Remote Sensing Magazine. 2017. P. 8-36.
- Levchenko D.A., Bojko E.S. Generator tsifrovyh dannyh geotochek iskusstvennyh territorij «Terra_Maker» [Digital geocode data generator for artificial territories "Terra_Maker"] // svidetel'stvo o gosudarstvennoj registratsii programmy dlja EVM, №2023610228
- Kolesnikov A.A. Analiz metodov i sredstv iskusstvennogo intellekta dlja analiza i interpretatsii dannyh aktivnogo distantsionnogo zondirovanija [Analysis of artificial intelligence methods and tools for the analysis and interpretation of active remote sensing data] // Vestnik SGUGiT (Sibirskogo gosudarstvennogo universiteta geosistem i tehnologij) [Bulletin of SGUGiT (Siberian State University of Geosystems and Technologies)]. 2022. V. 27. No 3. P. 74-94.
- Glinskij A. V., Novichenko E. V., Tkachev M. S., Savenkov V. M., Tepljakova A. R., Zaharochkina N. A., Birjukov R. N., Akobjan L. V. Programma semanticheskoj segmentatsii s uchetom dannyh, poluchaemyh s kamer i lidarov [Semantic segmentation program based on data received from cameras and lidars] // Svidetel'stvo o registratsii programmy dlja `EVM 2022663627, 18.07.2022.
- Ali Kashefi, Tapan Mukerji Physics-informed PointNet: A deep learning solver for steady-state incompressible flows and thermal fields on multiple sets of irregular geometries // Journal of Computational Physics, 2022
- D'jachenko R.A., Kosolapov P.A., Gura D.A. K voprosu ob uvelichenii proizvoditel'nosti mashinnogo obuchenija na`etape vyborki dannyh pri reshenii zadach klassifikatsii [On the issue of increasing machine learning performance at the data sampling stage when solving classification problems] // Vestnik Voronezhskogo gosudarstvennogo universiteta. Serija: Sistemnyj analiz i informatsionnye tehnologii [Bulletin of the Voronezh State University. Series: System analysis and Information Technology]. 2022. No 4. P. 146-155.9.
- Gura D.A., D'jachenko R.A., Andrjuschenko A.V. Tehnologija imitatsii poletnogo marshruta bespilotnogo letatel'nogo apparata i vozdushnogo lazernogo skanirovanija [Technology for simulation of the flight route of an unmanned aerial vehicle and aerial laser scanning] // `Elektronnyj setevoj politematicheskij zhurnal "Nauchnye trudy KubGTU" [Electronic network polythematic journal "Scientific works of KubSTU"]. 2022. No 6. P. 126-133
- Kosolapov P.A., D'jachenko R.A., Gura D.A., Husht N.I. K voprosu o normalizatsii dannyh trehmernogo lazernogo skanirovanija [On the issue of normalization of three-dimensional laser scanning data] // `Elektronnyj setevoj politematicheskij zhurnal "Nauchnye trudy KubGTU" [Electronic network polythematic journal "Scientific works of KubSTU"]. 2021. No 4. P. 56-66.
- Gura D.A., D'jachenko R.A., Husht N.I. K voprosu o predobrabotke dannyh trehmernogo lazernogo skanirovanija [On the issue of preprocessing three-dimensional laser scanning data] // `Elektronnyj setevoj politematicheskij zhurnal "Nauchnye trudy KubGTU" [Electronic network polythematic journal "Scientific works of KubSTU"]. 2021. N 3. P. 39-46.
- Gura D.A., D'jachenko R.A., Husht N.I., Markovskij I.G. K voprosu o metodah generatsii, klassifikatsii i raspoznavanija trehmernyh izobrazhenij [On the issue of methods of generation, classification and recognition of three-dimensional images] // `Elektronnyj setevoj politematicheskij zhurnal "Nauchnye trudy KubGTU" [Electronic network polythematic journal "Scientific works of KubSTU"]’. 2021. No 2. P. 86-97.
- Aoki Yu., Gofort H., Srivatsan R. A., Lucy S. Pointnetlk: Reliable and efficient point cloud registration using pointnet // Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 2019. P. 7156-7165.
- Zaganidis A., San L., Duckett T., Tselnyak G. Integration of deep semantic segmentation into three–dimensional point cloud registration // IEEE Robotics and Automation Letters. 2018. V. 3, No 4. P. 2942-2949.
- Zhang B., Huang S., Shen V., Wei Z. PointNet Explanation: What has been studied inside PointNet? // IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). 2019. P. 71-74.
Supplementary files
Supplementary Files
Action
1.
JATS XML
2.
Fig. 1. Architecture of the Paint net network
Download (251KB)
3.
Fig. 2. On-screen form of the Terra_Maker information system settings dialog "Ground surface"
Download (103KB)
4.
Fig. 3. On-screen form of the Terra_Maker information system settings dialog
Download (95KB)
5.
Fig. 4. On-screen form of the Terra_Maker information system settings dialog "Structures/Buildings"
Download (98KB)
6.
Fig. 5. Screen form of the Terra_Maker information system settings dialog "Falsely reflected points"
Download (27KB)
7.
Fig. 6. On-screen form of the Terra_Maker information system settings dialog "Numerical characteristics of the data set being created"
Download (33KB)
8.
Fig. 7. Visualization of point clouds in the Terra_Maker information system
Download (30KB)
9.
Figure 8. Visual representation of the point cloud dataset
Download (44KB)
10.
Fig. 9. Graphs of the values of the loss function and accuracy by training epochs
Download (20KB)
11.
Fig. 10. Graphs of the dependencies of the accuracy of training and testing
Download (20KB)
12.
Fig. 11. An example of semantic segmentation of laser reflection points_1
Download (10KB)
13.
Fig. 11. An example of semantic segmentation of laser reflection points_2
Download (10KB)
14.
Fig. 11. An example of semantic segmentation of laser reflection points_3
Download (9KB)
15.
Fig. 11. An example of semantic segmentation of laser reflection points_4
Download (9KB)
16.
Fig. 11. An example of semantic segmentation of laser reflection points_5
Download (9KB)
17.
Fig. 11. An example of semantic segmentation of laser reflection points_6
Download (10KB)
18.
Fig. 11. An example of semantic segmentation of laser reflection points_7
Download (8KB)
19.
Fig. 11. An example of semantic segmentation of laser reflection points_8
Download (8KB)
