AI-powered visual classification in archives: a computer vision approach to facial recognition in historical archives

This study examines the integration of computer vision technologies, specifically the YOLOv9 algorithm, into the management of historical visual archives, focusing on facial recognition to improve the organization, classification, and accessibility of extensive collections of photographs and videos. A curated dataset of 1,638 images of prominent cinema figures was expanded to 3939 images through data augmentation, and the YOLOv9 model was trained using preprocessing, annotation, and Google Colab’s GPU. The model demonstrated robust performance, achieving precision of 91.8%, recall of 85.2%, mAP50 of 93.4%, and mAP50-95 of 62.6%, showcasing its capability to handle large datasets with high accuracy. These findings highlight the transformative potential of computer vision in archival management, enabling more accessible and searchable visual materials. By extending its application to both static images and video content, this study contributes to archival science through the innovative use of advanced facial recognition techniques, offering a dynamic solution for modern archival systems.