Awesome

Video Face Clustering (ICCV 2019)

<strong>Video Face Clustering with Unknown Number of Clusters</strong>
M. Tapaswi, M. T. Law, and S. Fidler
International Conference on Computer Vision (ICCV), October 2019.
arXiv

• Realistic setting for clustering face tracks in videos
• Number of clusters is not known
• Background character face tracks are not removed and need to be resolved
• <strong>Ball Cluster Learning</strong>: a new loss function that carves feature space into balls of a learned radius that can be used as a stopping criterion of agglomerative clustering

Dataset

We use 6 episodes of season 1 of <em>The Big Bang Theory</em> and 6 episodes of season 5 of <em>Buffy - The Vampire Slayer</em>. Face track labels are resolved between background characters.

Original tracks were provided by:
M. Bäuml, et al. Semi-supervised Learning with Constraints for Person Identification in Multimedia Data. CVPR 2013.

Please use the <code>download.sh</code> script inside <code>data/</code> for convenience.
Face tracks (5.3 MB)
VGG Face SE-ResNet50-256 features (519 MB)

Evaluation

(02.09.2019) Our final checkpoint with evaluation code has been released. Numbers can be reproduced by downloading the features and labels and running with video_name as one of the following videos: bbt_s01e01..06 or buffy_s05e01..06

python evaluate.py <video_name>

Code

coming soon