Year: 2025
Type: article
Abstract: The proliferation of counterfeit logos necessitates efficient methods for brand protection. This study addresses this challenge by introducing Orifake, a deep-learning model leveraging YOLOv8 (You Only Look Once) for logo classification. The model analyses logos (Adidas, Nike, Puma) for subtle features like colour, patterns, shapes, and textures. The evaluation demonstrates an average recall of 78.5%. Notably, Orifake excels at identifying fake NIKE logos (recall: 94.9%). However, further refinement is needed for original NIKE logos due to their lower mAP-95 value. This research highlights the potential of Artificial Intelligence (AI) logo detection for brand protection against increasing online fraud.
Institutions: Universiti Teknologi MARA System, Universiti Teknologi MARA
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Funder: Universiti Teknologi MARA
OpenAlex ID: W4412188020