EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

A Patch-Based Voting Framework with MobileNetV2 for Identifying AI-Generated Visual Content in Digital Poster

Cintha Hafrida Putri

Journal of Artificial Intelligence and Data Science · 2025

DOI: 10.67561/jaids.v1i2.5 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Artificial Intelligence and Data Science, 2025. Verfasst von Cintha Hafrida Putri. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study aims to detect the involvement of the use of AI in digital posters. The algorithm used for this purpose is to use the MobileNetV2 network architecture with evaluation scenarios to determine the best model configuration on the detection of the use of AI involved in digital posters. In this study, a method is proposed that uses a patch-based voting mechanism to model local visual patterns that serve as evidence of creator presence. The dataset used totaled 208 posters, half of which were man-made and the other half were AI-generated. The proposed model based on the CRISP-DM framework consists of six steps: (1) business understanding; (2) understanding data; (3) data preparation; (4) modeling; (5) evaluation; (6) Application. Various permutations of MobileNetV2 training including learning rate tuning, unlocked layers, and enriched data are explored to find the most reproducible architecture with the highest performance. The best models are selected based on their performance on validation data and voting uniformity across patches. The results of the model showed a training accuracy of 94.74%, a validation accuracy of 91.18%, and a test accuracy of 90.48% with a strong ability to distinguish visual features between AI and human work. These results suggest that the selection of appropriate training cases for MobileNetV2 along with a patch-based approach is a good way to filter the influence of AI in contemporary visual content.

Zitieren

APA

Putri, C. H. (2025). A Patch-Based Voting Framework with MobileNetV2 for Identifying AI-Generated Visual Content in Digital Poster. Journal of Artificial Intelligence and Data Science. https://doi.org/10.67561/jaids.v1i2.5

Vancouver

Putri CH. A Patch-Based Voting Framework with MobileNetV2 for Identifying AI-Generated Visual Content in Digital Poster. Journal of Artificial Intelligence and Data Science. 2025. doi:10.67561/jaids.v1i2.5

BibTeX
@article{Putri2025,
  author = {Putri, Cintha Hafrida},
  title = {A Patch-Based Voting Framework with MobileNetV2 for Identifying AI-Generated Visual Content in Digital Poster},
  journal = {Journal of Artificial Intelligence and Data Science},
  year = {2025},
  doi = {10.67561/jaids.v1i2.5}
}
Kostenlos starten →