ANALYSIS OF SOTA-ARCHITECTURES FOR SEMANTIC AUGMENTATION

Ye. S. Moroz, Y.S. Pikaliov

About the authors

Ye. S. Moroz, Y.S. Pikaliov

Мороз Егор Сергеевич – инженер-исследователь, Федеральное государствен- ное бюджетное научное учреждение, «Институт проблем искусственного интеллекта». Область научных интересов: анализ данных, распознавание образов, компьютерное зрение, адрес: г. Донецк, ул. Егорова, д. 16, кв. 1.

Пикалёв Ярослав Сергеевич – кандидат техн. наук, старший научный сотрудник лаборатории интеллектуальных систем и анализа данных Федерального государственного бюджетного научного учреждения «Институт проблем искусственного интеллекта», Область научных интересов: цифровая обработка сигналов, анализ данных, распознавание образов, обработка естественного языка, компьютерное зрение, машинное обучение, нейронные сети, г. Донецк, адрес: 283085, ДНР, г. Донецк, ул. Отважных, д. 19, кв. 85.

Download PDF

UDC: 004.932.2
DOI: 10.24412/2413-7383-2025-4-39-60-70
Language: Russian

Abstract: The paper considers semantic segmentation tasks in the context of applied computer vision and related problems such as object detection, instance segmentation, and panoptic segmentation. Lightweight and high-speed real-time architectures are analyzed in terms of the trade-off between accuracy and latency on the Cityscapes, CamVid, and UAVid datasets. Special attention is paid to data preparation issues, including augmentation, image annotation, preprocessing, and model interpretability.

Keywords: semantic segmentation, computer vision, object detection, instance segmentation, panoptic segmentation, data augmentation, cross-view geolocalization.

References

1. Moroz YE. S. Methods of semantic segmentation in computer vision: a review of architectures, loss functions and modern approaches / E. S. Moroz // Donetsk Readings – 2025: education, science, innovation, culture and challenges of the present: materials of the 10th International scientific conference (Donetsk, 2025). – Donetsk, 2025.
2. Pokintelitsa A. E. Application of cellular automata for creating artificial digital images / A. E. Pokintelitsa // Artificial Intelligence: theoretical aspects and practical application: proceedings of the Donetsk International Round Table. – Donetsk: FSBSI “Institute for Problems of Artificial Intelligence”, 2025. –
3. p. – P. 12–18.
4. Ermolienko T. V. On the application of deep learning to the problem of cross-view geolocalization / T. V. Ermolienko, R. S. Khakimov // Problems of Artificial Intelligence. – 2024. – No. 4 (35). – P. 12–28.
5. Pikalev, Ya. S. Detection of key objects and cross-view geolocation: analysis of datasets and methodological aspects / Ya. S. Pikalev. // Problems of Artificial Intelligence. – 2024. – No. 4. – P. 25–37. – ISSN 2413-7383.
6. Pavlenko, B. V. Pikalev, Ya. S. Methodology for creating a set of aerial images for the cross-view geolocation problem / B. V. Pavlenko, Ya. S. Pikalev. // Problems of Artificial Intelligence. – 2024. – No.
7. – P. 101–112. – ISSN 2413-7383.
8. Blizno M. V. Comparison of various augmentation methods for training a neural network model / M. V. Blizno // Artificial Intelligence: theoretical aspects and practical application: proceedings of the Donetsk International Round Table. – Donetsk: FSBSI “Institute for Problems of Artificial Intelligence”, 2025. –
9. p. – P. 99–102.
10. Khakimov R. S. On the development of a data annotation system for computer vision tasks / R. S. Khakimov, O. L. Nizhnikova, M. V. Blizno // Problems of Artificial Intelligence. – 2024. – No. 3 (34). – P. 70–79. – ISSN 2413-7383. – DOI 10.24412/2413-7383-2024-3-70-79.
11. Akimov A. A. Preprocessing of data for machine learning / A. A. Akimov, D. R. Valitov, A. I. Kubryak //Scientific Review. Engineering Sciences. – 2022. – No. 2. – P. 26–31.
12. Krishnan S. R. Improving anomaly detection in video using an enhanced UNET technology and a cascaded sliding-window technique / S. R. Krishnan, P. Amudha // Informatics and Automation. – 2024. – No. 6 (23). – P. 1899–1930. – ISSN 2713-3192. – DOI 10.15622/ia.23.6.12.
13. Khakimov R. S. Review of advanced augmentation techniques for image datasets / R. S. Khakimov, B. V. Pavlenko, Ya. S. Pikalev // Donetsk Readings 2024: materials of the 9th International scientific conference (Donetsk, 15–17 October 2024). – Vol. 2. – Donetsk: DonNU Press, 2024. – P. 272–275. – ISSN 2664- 7362 (Print); ISSN 2664-7370.
14. Nikitenko, K. A. Zvyagintseva, A. V. Interpretability of neurosemantic models in their application to practical domains / K. A. Nikitenko, A. V. Zvyagintseva. // Problems of Artificial Intelligence. – 2025. – No. 2. – P. 79–90. – ISSN 2413-7383.
15. Solopov, M. V. Chechekhina, E. S. Popandopulo, A. G. Kavelina, A. S. Akopyan, G. V. Turchin, V. V. Study of the application of Mask R-CNN and Segment Anything Model (SAM) for instance segmentation of mesenchymal stem cells in microphotographs / M. V. Solopov, E. S. Chechekhina, A. G. Popandopulo, Анализ Sota-архитектур для семантической аугментации М A. S. Kavelina, G. V. Akopyan, V. V. Turchin. // Problems of Artificial Intelligence. – 2025. – No. 2(37). – P. 21–29. – DOI: 10.24412/2413-7383-2025-2-37-21-29. – EDN: ODOKND. – ISSN 2413-7383.
16. Yu C. BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation / C. Yu, J. Wang, C. Peng, C. Gao, G. Yu, N. Sang // Proc. ECCV. – 2018. – Electronic resource. – Mode of access: https://www.alphaxiv.org/abs/1808.00897
17. Xu J. PIDNet: A Real-time Semantic Segmentation Network Inspired by PID Controllers / J. Xu, Y. Tang, K. Chen [et al.]. arXiv preprint, 2022. Electronic resource. Mode of access: https://arxiv.org/abs/2206.02066
18. Yin H. Deep Dual-resolution Networks for Real-time and Accurate Semantic Segmentation of Road Scenes / H. Yin, X. Shen, J. Fang [et al.]. arXiv preprint, 2021. Electronic resource. Mode of access: https://arxiv.org/pdf/2101.06085
19. Xie E. SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers / E. Xie, W. Wang, Z. Yu [et al.]. arXiv preprint, 2024. Electronic resource. Mode of access: https://arxiv.org/abs/2410.01092v1
20. Li H. DFNet: Deep Feature Aggregation Network for Real-time Semantic Segmentation / H. Li, P. Xiong, J. Fan, L. Sun. – arXiv preprint, 2019. – Electronic resource. – Mode of access: https://arxiv.org/abs/1903.03777v2
21. Li X. Semantic Flow for Fast and Accurate Scene Parsing (SFNet) / X. Li, H. Xiong, H. Fan [et al.] // Proc. ECCV. – 2020. – Electronic resource. – Mode of access: https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123460749.pdf
22. Yu C. BiSeNet V2: Bilateral Network with Guided Aggregation for Real-time Semantic Segmentation / C. Yu, C. Gao, J. Wang, G. Yu, N. Sang. – arXiv preprint, 2020. – Electronic resource. – Mode of access: https://arxiv.org/pdf/2004.02147
23. Chen X. FasterSeg: Searching for Faster Real-time Semantic Segmentation / X. Chen, Y. Zhu, G. Lin [et al.]. – arXiv preprint, 2019. – Electronic resource. – Mode of access: https://arxiv.org/abs/1912.10917
24. When Humans Meet Machines: Towards Efficient Segmentation / [no authors]. – [s.l.], 2023. – Electronic resource. – Mode of access: https://scispace.com/pdf/when-humans-meet-machines-towards-efficient- segmentation-4pir2uvpjc.pdf
25. Fan M. Rethinking BiSeNet for Real-time Semantic Segmentation (STDC-Seg) / M. Fan, T. Wang, D.

Issue: 4(39)'2025
Section: ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING
How to cite: Ye. S. Moroz, Y.S. Pikaliov. ANALYSIS OF SOTA-ARCHITECTURES FOR SEMANTIC AUGMENTATION // Problems of Artificial Intelligence. - 2025. - № 4 (39). - P. 60-70. - https://paijournal.guiaidn.ru/en/2025/4(39)-6.html

Back to issue 4(39)'2025