Russian Federation
Russian Federation
The paper presents a formalized model of the image generation process within two architectural paradigms: generative adversarial networks and diffusion-based models. A quantitative comparison is conducted using inference step count and FLOPs estimation, accompanied by a qualitative analysis of training stability, mode collapse susceptibility, latent space interpretability, and scalability.
generative models; GAN; diffusion models; computational complexity; FID; process modeling; image generation
1. Yurchishina, M. V. Algoritmicheskaya model' SPPR «Optimal'nyy uchebnyy plan» / M. V. Yurchishina, K. I. Bushmeleva // Modelirovanie sistem i processov. – 2024. – T. 17, № 4. – S. 84-95. – DOIhttps://doi.org/10.12737/2219-0767-2024-17-4-84-95. – EDN OJFBGD.
2. Rombah R., Blattman A., Lorenc D., Esser P., Ommer B. Vysokokachestvennyy sintez izobrazheniy s ispol'zovaniem latentnyh diffuzionnyh modeley // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2022.
3. Sahariya Ch., Chan U., Saksena S., Li L., Uitaker S., Barrett U., i dr. Fotorealistichnaya generaciya teksta v izobrazhenie s glubokim ponimaniem yazyka (Imagen) // arXiv. 2022.
4. Diffusion Models vs GANs vs VAEs: Comparison of Deep Generative Models // Towards AI. URL: https://pub.towardsai.net/diffusion-models-vs-gans-vs-vaes-comparison-of-deep-generative-models-67ab93e0d9ae (data obrascheniya: 10.02.2026).
5. Vivekananthan S. Sravnitel'nyy analiz generativnyh modeley: povyshenie kachestva sinteza izobrazheniy s ispol'zovaniem VAE, GAN i Stable Diffusion // arXiv. 2024.
6. Song Ya., Dhariual P., Chen M., Sutskever I. Consistency Models // International Conference on Machine Learning (ICML). 2023.
