An evaluation of lightweight facial recognition models available in the DeepFace library in clustering tasks
Year: 2026
Type: conference-paper
Abstract: This work investigates the use of Deep Learning-based facial recognition models to automatically organize images through unsupervised clustering. We evaluated the FaceNet and ArcFace models using the DeepFace library on a dataset of 100 images of five celebrities. The results show an average accuracy of 92.6%, demonstrating that the choice of model and the clustering algorithm configuration significantly influence performance and enable high accuracy in automatic image organization.
Institution: Universidade Tecnológica Federal do Paraná
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7204201393
