Controlled Image Transformations for Expanding AI Model Training Sets in Computer Vision

Scientific adviser:
Alexander A. Rogov
Reviewer:
Dmitry G. Korzun

Paper submitted on: 05/17/2026;
Accepted on: 06/27/2026;
Published online on: 06/27/2026.

Abstract. Manual preparation of images for computer vision tasks is labor-intensive and expensive, limiting an AI model's training set and leading to overfitting and instability under varying acquisition conditions. This work systematizes methods of controlled image transformations that augment the training set and examines their impact on model quality and security. The proposed approach improves accuracy, generalization ability, and robustness to adversarial attacks. The article also discusses transitioning to generative synthesis of complex noise, which enables the development of reliable systems operating under rare distortions or phenomena.
Keywords: image augmentation, controlled transformations, training set, model robustness, adversarial defense, generative models

For citation: Pavlov, M. P. Controlled Image Transformations for Expanding AI Model Training Sets in Computer Vision. StudArctic forum. 2026, 11 (2): 24–32.

References

Vorona A.A., Sevastei E.A. Methods for increasing robustness of neural networks to adversarial attacks in computer vision systems. Molodoy uchyoniy, 2024, No. 7(610), pp. 3–8. (In Russ.)

Gerasimov V.M., Maslova M.A., et al. Protection against adversarial attacks on audio and images in artificial intelligence models using the SGEC method. Research Result. Information Technologies, 2023, Vol. 8, No. 2, pp. 53–60. DOI: 10.18413/2518-1092-2022-8-2-0-7 (In Russ.)

Chekhonina E.A., Kostyumov V.V. Overeview of adversarial attacks and defenses for object detectors. International Journal of Open Information Technologies, 2023, No. 7, pp. 11-20. (In Russ.)

Athalye C., Arnaout R. Domain-guided data augmentation for deep learning on medical imaging. PLOS ONE, 2023, Vol. 18, No. 3. DOI: 10.1371/journal.pone.0282532

Buslaev A., Iglovikov V., et al. Albumentations: Fast and flexible image augmentations. Information, 2020, Vol. 11, No. 2, p. 125. DOI: 10.3390/info11020125

Cubuk E.D., Zoph B., et al. AutoAugment: Learning augmentation policies from data. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019, pp. 113–123. DOI: 10.1109/CVPR.2019.00020

Cubuk E.D., Zoph B., et al. RandAugment: Practical automated data augmentation with a reduced search space. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Seattle, WA, USA, 2020, pp. 3008–3017. DOI: 10.1109/CVPRW50498.2020.00359

DeVries T., Taylor G.W. Improved regularization of convolutional neural networks with cutout. arXiv preprint. Cornell University, 2017. DOI: 10.48550/arXiv.1708.04552

Dhariwal P., Nichol A. Diffusion models beat GANs on image synthesis. Advances in Neural Information Processing Systems (NeurIPS), 2021, Vol. 34, pp. 8780–8794.

Goodfellow I.J., Shlens J., et al. Explaining and harnessing adversarial examples. Proceedings of the 3rd International Conference on Learning Representations (ICLR). San Diego, CA, USA, 2015.

Li L., Qiu J., et al. AROID: Improving adversarial robustness through online instance-wise data augmentation. International Journal of Computer Vision, 2024, Vol. 132, No. 2, pp. 929–950. DOI: 10.1007/s11263-023-01912-9

Madry A., Makelov A., et al. Towards deep learning models resistant to adversarial attacks. Proceedings of the 6th International Conference on Learning Representations (ICLR). Vancouver, BC, Canada, 2018. DOI: 10.48550/arXiv.1706.06083

Shorten C., Khoshgoftaar T.M. A survey on Image Data Augmentation for Deep Learning. Journal of Big Data, 2019, Vol. 6, No. 1, p. 60. DOI: 10.1186/s40537-019-0197-0

Yun S., Han D., et al. CutMix: Regularization strategy to train strong classifiers with localizable features. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). 2019, pp. 6023–6032. DOI: 10.1109/ICCV.2019.00612

Zhang H., Cisse M., et al. Mixup: Beyond empirical risk minimization. Proceedings of the 6th International Conference on Learning Representations (ICLR). Vancouver, BC, Canada, 2018. DOI: 10.48550/arXiv.1710.09412

Zhu J.-Y., Park T., et al. Unpaired image-to-image translation using cycle-consistent adversarial networks. Proceedings of the IEEE International Conference on Computer Vision (ICCV). 2017, pp. 2223–2232. DOI: 10.1109/ICCV.2017.244

Displays: 185; Downloads: 63;

StudArctic Forum

scientific journal

Controlled Image Transformations for Expanding AI Model Training Sets in Computer Vision

References