Do We Really Need to Collect Millions of Faces for Effective Face Recognition?

TL;DR

This paper explores synthesizing facial images to augment training data for face recognition, potentially reducing the need for collecting millions of real images, and achieves comparable state-of-the-art results.

Contribution

It introduces novel face synthesis methods to enrich datasets, demonstrating effective training without extensive real image collection.

Findings

Synthesized images improve face recognition accuracy.

Performance matches state-of-the-art systems trained on large datasets.

Effective augmentation reduces reliance on massive data collection.

Abstract

Face recognition capabilities have recently made extraordinary leaps. Though this progress is at least partially due to ballooning training set sizes -- huge numbers of face images downloaded and labeled for identity -- it is not clear if the formidable task of collecting so many images is truly necessary. We propose a far more accessible means of increasing training data sizes for face recognition systems. Rather than manually harvesting and labeling more faces, we simply synthesize them. We describe novel methods of enriching an existing dataset with important facial appearance variations by manipulating the faces it contains. We further apply this synthesis approach when matching query images represented using a standard convolutional neural network. The effect of training and testing with synthesized images is extensively tested on the LFW and IJB-A (verification and identification) benchmarks and Janus CS2. The performances obtained by our approach match state of the art results reported by systems trained on millions of downloaded images.