Homepage Program Systems: Theory and Applications Русская версия
ISSN 2079-3316 Bilingual online scientific Online scientific journal of the Ailamazyan Program System Institute of the Ailamazyan PSI of PSI of Russian Academy of Science of RAS 12+ 
Volume 16 (2025) . Issue 2 (65) . Paper No. 4 (451)

Artificial intelligence and machine learning

Research Article
DOIDigital Object Identifier 10.25209/2079-3316-2025-16-2-111-152

Improving the accuracy of segmentation masks using a generative-adversarial network model

Igor Victorovich VinokurovCorrespondent author

Financial University under the Government of the Russian Federation, Moscow, Russia
Igor Victorovich Vinokurov — Correspondent author igvvinokurov@fa.ru

Abstract. Masks obtained using the deep learning model Mask R-CNN may in some cases contain fragmented contours, uneven boundaries, false fusions of adjacent objects, and areas with missed segmentation. The more detection objects in the image and the smaller their size, the more often various types of defects in their masks are encountered. Examples of such images include aerial photographs of cottage and garden associations and cooperatives characterized by high building density. To correct these defects, it is proposed to use a generative adversarial network model that performs post-processing of the predicted Mask R-CNN masks.

A qualitative assessment of the model formed in the work demonstrated that it is capable of restoring the integrity of contours at an acceptable level, filling in missing areas, and separating erroneously merged objects. Quantitative analysis using the IoU, precision, recall, and F1-score metrics showed a statistically significant improvement in the segmentation quality compared to the original Mask R-CNN masks. The obtained results confirmed that the proposed approach allows to increase the accuracy of the formation of object masks to a level that satisfies the requirements of their practical application in automated aerial photograph analysis systems. (Linked article texts in English and in Russian).

Keywords: Computer vision, image segmentation, object masks, generative adversarial networks, Mask R-CNN, PyTorch

MSC-20202020 Mathematics Subject Classification 68T20; 68T07, 68T45MSC-2020 68-XX: Computer science MSC-2020 68Txx: Artificial intelligence MSC-2020 68T20: Problem solving in the context of artificial intelligence (heuristics, search strategies, etc.) MSC-2020 68T07: Artificial neural networks and deep learning MSC-2020 68T45: Machine vision and scene understanding

For citation: Igor V. Vinokurov. Improving the accuracy of segmentation masks using a generative-adversarial network model. Program Systems: Theory and Applications, 2025, 16:2, pp. 111–152. (in Engl. In Russ.). https://psta.psiras.ru/2025/2_111-152.

Full text of bilingual article (PDF): https://psta.psiras.ru/read/psta2025_2_111-152.pdf (Clicking on the flag in the header switches the page language).

The article was submitted 21.04.2025; approved after reviewing 10.06.2025; accepted for publication 11.06.2025; published online 28.06.2025.

© Vinokurov I. V.
2025
Editorial address: Ailamazyan Program Systems Institute of the Russian Academy of Sciences, Peter the First Street 4«a», Veskovo village, Pereslavl area, Yaroslavl region, 152021 Russia;   Website:  http://psta.psiras.ru Phone: +7(4852) 695-228;   E-mail: ;   License: CC-BY-4.0License text on the Creative Commons site
© Ailamazyan Program System Institute of Russian Academy of Science (site design) 2010–2025