EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI-based painting style classification: promoting creativity development through fine art education using GS-Swin Transformer

Yuxin Jiang, Lingjie Chen, Chenyang Guo, Guobin Sun

PeerJ Computer Science · 26.05.2026 · 12 · e3829

DOI: 10.7717/peerj-cs.3829 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PeerJ Computer Science, online erschienen am 26.05.2026. Verfasst von Yuxin Jiang und 3 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Effective image literacy is an important approach to cultivating children’s creativity through fine art education. However, due to the unequal distribution of educational resources, fine art teachers in some regions are unable to provide high-quality image literacy instruction, which hinders the cultivation of children’s creativity. Artificial intelligence-driven transformations in the field of education have provided new ideas for addressing this issue. Therefore, this article constructs the GS-Swin Transformer–based painting style intelligent recognition model and conducts its training and evaluation on a self-built dataset aligned with the theme of fine art education. The results show that the GS-Swin Transformer achieved a Top-1 classification performance on the test set, with Accuracy, Precision, Recall, and F1-score reaching 85.56%, 85.67%, 85.56%, and 85.60%, respectively. Compared with Swin Transformer and Vision Transformer, the proposed model shows improvements of approximately 1.85% and 4.30% in the evaluation metrics, respectively. Compared with the three convolutional neural network (CNN)-based baseline comparison models, the improvements range from approximately 8.44% to 27.02%. Compared with GS-Vision Transformer, the improvement of evaluation metrics is still approximately 3.06%. Compared with grid-search-optimized CNN baseline comparison models, evaluation metric still can be improved from approximately 7.66% to 18.00%. The model demonstrates excellent learning efficiency, generalization ability, and performance, which highlights the advantages of the Swin Transformer architecture and confirms the effectiveness of the grid search optimization strategy. On this basis, a human-artificial intelligence (AI) collaborative intelligent image literacy instruction model is proposed to assist fine art teachers in effectively expanding their teaching resources and providing high-quality image literacy instruction for students. This provides a potential practical pathway to narrow the experience gap among fine art teachers in different areas, foster the development of children’s creativity, and enhance the application of AI in fine art education.

Zitieren

APA

Jiang, Y., Chen, L., Guo, C., & Sun, G. (2026). AI-based painting style classification: promoting creativity development through fine art education using GS-Swin Transformer. PeerJ Computer Science, 12, e3829. https://doi.org/10.7717/peerj-cs.3829

Vancouver

Jiang Y, Chen L, Guo C, Sun G. AI-based painting style classification: promoting creativity development through fine art education using GS-Swin Transformer. PeerJ Computer Science. 2026;12:e3829. doi:10.7717/peerj-cs.3829

BibTeX
@article{Jiang2026,
  author = {Jiang, Yuxin and Chen, Lingjie and Guo, Chenyang and Sun, Guobin},
  title = {AI-based painting style classification: promoting creativity development through fine art education using GS-Swin Transformer},
  journal = {PeerJ Computer Science},
  year = {2026},
  volume = {12},
  pages = {e3829},
  doi = {10.7717/peerj-cs.3829}
}
Kostenlos starten →