Deep Face Recognition

Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman
2015 Procedings of the British Machine Vision Conference 2015  
The goal of this paper is face recognition -from either a single photograph or from a set of faces tracked in a video. Recent progress in this area has been due to two factors: (i) end to end learning for the task using convolutional neural networks (CNNs), and (ii) the availability of very large scale training datasets. We make two contributions: first, we show how a very large scale dataset (2.6M images spanning more than 2.6K identities) can be constructed by semi-automatic annotations with
more » ... umans in the loop, investigating the trade-off between annotation purity and cost; second, we introduce a very deep convolutional neural network and a corresponding training procedure that achieve face recognition accuracy comparable to the current state of the art on public benchmarks such as "Labelled Faces In the Wild" and "YouTube Faces Dataset", while at the same time using a fraction of the data used by competitors. Figure 1: Example images from our dataset.
doi:10.5244/c.29.41 dblp:conf/bmvc/ParkhiVZ15 fatcat:vhncyzl2dbd5fjmf6rama7oqia