← 返回论文检索
ACM Multimedia 2025Demos and Videos

FaceCluster: Interactive Photo Organization with Enhanced Face Recognition

Alexander Filonenko, Ilya Makarov, Andrey V. Savchenko

PDF 由论文原始站点提供,PaperCompass 不保存论文文件。DOI 10.1145/3746027.3754485 ↗

摘要

FaceCluster is an interactive photo management system that leverages our enhanced KP-RPE face recognition model with Embedding Statistical Regularization to organize personal photo collections automatically. Unlike existing cloud-based systems that raise privacy concerns, FaceCluster operates entirely locally while demonstrating high performance across multiple challenging benchmarks, including IJB-C (97.25% TAR@0.01%), TinyFace (74.14% Rank-1), and AgeDB (97.78% accuracy) when trained on the WebFace4M dataset. The demo showcases real-time face detection, clustering, and organization capabilities through an intuitive web interface, enabling users to effortlessly manage large photo collections with a single-command Docker deployment.