EUVIMEDEUVIMED entdecken
Crossref

Mask-Aware Localized Inpainting Method for CPU-Based Inference

Volodymyr Oliinyk, Serhii Hatsan

International Journal of Engineering and Manufacturing · 08.06.2026 · 16 (3) · 30-41

DOI: 10.5815/ijem.2026.03.03 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Engineering and Manufacturing, erschienen am 08.06.2026. Verfasst von Volodymyr Oliinyk und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Image generation methods, including inpainting, are evolving rapidly; however, high memory requirementscontinue to limit their practical deployment. As a result, the efficient utilization of Latent Diffusion Models onedgedevices has become increasingly important. This work explores techniques for reducing memory usage inLatent Diffusion Models while preserving their generative capabilities. We propose a resource-efficient inpainting method optimized for CPU-based inference, based on a combinationof VAEtiling, attention slicing, and dynamic region-of-interest slicing. Experimental results demonstrate that the model's memoryfootprint can be significantly reduced while maintaining output quality, without substantial increases in computationtime, enabling execution on systems with as little as 4 GB of memory and only two processing cores. While the introducedoptimizations, particularly those based on localized image processing, introduce an inherent trade-off betweenmemoryusage and computational cost, resulting in longer inference times compared to GPU-accelerated solutions, theydemonstrate strong potential for deployment in memory-limited environments. Additionally, we provide analysis of key deployment bottlenecks, including model compilation for cold-start overheadmitigation, proper runtime configuration and scheduler selection. These findings confirm the feasibility of effectivelydeploying Latent Diffusion Models for inpainting tasks on CPU-only, resource-constrained platforms, thereby broadeningtheir applicability to edge computing scenarios

Zitieren

APA

Oliinyk, V., & Hatsan, S. (2026). Mask-Aware Localized Inpainting Method for CPU-Based Inference. International Journal of Engineering and Manufacturing, 16(3), 30-41. https://doi.org/10.5815/ijem.2026.03.03

Vancouver

Oliinyk V, Hatsan S. Mask-Aware Localized Inpainting Method for CPU-Based Inference. International Journal of Engineering and Manufacturing. 2026;16(3):30-41. doi:10.5815/ijem.2026.03.03

BibTeX
@article{Oliinyk2026,
  author = {Oliinyk, Volodymyr and Hatsan, Serhii},
  title = {Mask-Aware Localized Inpainting Method for CPU-Based Inference},
  journal = {International Journal of Engineering and Manufacturing},
  year = {2026},
  volume = {16},
  number = {3},
  pages = {30--41},
  doi = {10.5815/ijem.2026.03.03}
}
Kostenlos starten →