Friends Season 1

We present a novel face image dataset constructed from the first season of the popular TV series Friends, focusing on the main cast members. Face regions were automatically extracted using the Dlib library, resulting in a collection of high-quality facial images suitable for clustering and recognition tasks. This dataset is specifically designed to facilitate research in unsupervised face labeling, with a particular emphasis on spectral clustering methods. By leveraging the natural scene variability and frequent character co-occurrence in sitcoms, our dataset provides a realistic and challenging benchmark for evaluating the effectiveness of graph-based face clustering algorithms. We aim to support the development of more robust, scalable, and label-efficient face analysis techniques.