Journal of Radio Electronics. eISSN 1684-1719. 2025. ¹6

Contents

Full text in Russian (pdf)

Russian page

 

 

DOI: https://doi.org/10.30898/1684-1719.2025.6.2

 

 

 

Evaluation of the effectiveness of using
some SOTA neural network models in solving
classification problems of quasi-stationary objects
in the visible wavelength range

 

M.L. Belokopytov

 

Military space academy by A. Mozhaysky name
197198, Russia, St. Petersburg, Zhdanovskaya st. 13

 

The paper was received February 25, 2025.

 

Abstract. The paper provides a brief overview of modern neural network approaches to detecting quasi-stationary objects in optical images. The features of the construction, architecture and operation of five SOTA models of neural networks are considered. The data set was prepared (image parsing and markup, separation into training, validation and test samples). Neural network models such as Faster-CNN, DETR, RetinaNet, SSD, and YOLOv8m were trained and tested on the prepared dataset. A comparative analysis of trained neural networks was performed and the effectiveness of recognition algorithms was evaluated based on the accuracy of the models and their performance. The conclusion is made about the possibility of using such neural network detectors in solving the problem of automated processing of phono-target information.

Key words: neural network, dataset, sampling, training, metric, efficiency.

Corresponding author: Mark Lvovich Belokopytov, Hommer1990@mail.ru

References

1. Korchagin V.D. Analiz sovremennykh SOTA-arkhitektur iskusstvennykh nejronnykh setej dlya resheniya zadach klassifikacii izobrazhenij i detekcii ob"ektov // Programmnye sistemy i vychislitel'nye metody. – 2023. – ¹. 4. – S. 73-87. https://doi.org/10.7256/2454-0714.2023.4.69306.

2. Trask E. Grokaem glubokoe obuchenie. – Piter, 2024.

3. Chorbaa N.A., Tu L.A., Tolstoj I.M. Sravnitel'nyj analiz metodov detektirovaniya ob"ektov na radiolokacionnyk     h izobrazheniyakh pri pomoshchi nejronnykh setej // Nauchnyj rezul'tat. Informacionnye tekhnologii. – 2020. – T. 5. – ¹. 4. – S. 15-25. https://doi.org/10.18413/2518-1092-2020-5-4-0-3.

4. Carion N. et al. End-to-end object detection with transformers // European conference on computer vision. – Cham: Springer International Publishing, 2020. – P. 213-229.

5. Bolhovitina E.I. Issledovanie modelej svyortochnykh nejronnykh setej YOLOV3 i RETINANET dlya zadachi detektirovaniya lica cheloveka na izobrazhenii // StudNet. – 2022. – T. 5. – ¹ 6. – S. 5439-5448.

6. Liu W. et al. Ssd: Single shot multibox detector // Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14. – Springer International Publishing, 2016. – P. 21-37.

7. Brekht EH.A., Konshina V.N. Primenenie nejronnoj seti YOLO dlya raspoznavaniya defektov // Intellektual'nye tekhnologii na transporte. – 2022. –
¹ 2 (30). – S. 41-47. https://doi.org/10.24412/2413-2527-2022-230-41-47.

8. Timoshkin M.S., Mironov A.N., Leont'ev A.S. Sravnenie YOLO V5 i Faster R-CNN dlya obnaruzheniya lyudej na izobrazhenii v potokovom rezhime // Mezhdunarodnyj nauchno-issledovatel'skij zhurnal. – 2022. – ¹ 6-1 (120). – S. 137-146. https://doi.org/10.23670/IRJ.2022.120.6.020.

9. Sitnova A.V., Valitov EH.R., Svetozarskij S.N. Primenenie algoritmov glubokogo mashinnogo obucheniya na osnove mnogoslojnoj nejronnoj seti YOLOv8 dlya identifikacii gribkovogo keratita // Sovremennye tekhnologii v medicine. – 2024. – T. 16. – ¹ 4. – S. 5-14. https://doi.org/10.17691/stm2024.16.4.01.

10. Kopenkov V.N. Sovmestnyj analiz opticheskikh i radiolokacionnykh dannykh distancionnogo zondirovaniya Zemli v zadachakh analiza i monitoringa territorii // Sbornik trudov ITNT-2019. – 2019. – S. 21-24.

For citation:

Belokopytov M.L. Evaluation of the effectiveness of using some SOTA neural network models in solving classification problems of quasi-stationary objects in the visible wavelength range. // Journal of Radio Electronics. – 2025. – ¹6. https://doi.org/10.30898/1684-1719.2025.6.2 (In Russian)