Exploring Unlabeled Faces for Novel Attribute Discovery

TL;DR

This paper introduces a method for discovering new facial attributes and performing high-quality image translation using only unlabeled face images, reducing reliance on labeled datasets.

Contribution

It proposes a novel normalization technique guided by prior visual knowledge to discover attributes and enable translation without labeled data.

Findings

High-quality translations from unlabeled data

Preserves identity and realism in generated images

Outperforms some state-of-the-art methods trained on labeled data

Abstract

Despite remarkable success in unpaired image-to-image translation, existing systems still require a large amount of labeled images. This is a bottleneck for their real-world applications; in practice, a model trained on labeled CelebA dataset does not work well for test images from a different distribution -- greatly limiting their application to unlabeled images of a much larger quantity. In this paper, we attempt to alleviate this necessity for labeled data in the facial image translation domain. We aim to explore the degree to which you can discover novel attributes from unlabeled faces and perform high-quality translation. To this end, we use prior knowledge about the visual world as guidance to discover novel attributes and transfer them via a novel normalization method. Experiments show that our method trained on unlabeled data produces high-quality translations, preserves identity, and be perceptually realistic as good as, or better than, state-of-the-art methods trained on labeled data.