MatchGAN: A Self-Supervised Semi-Supervised Conditional Generative Adversarial Network
This work addresses the challenge of reducing labelled data requirements for conditional GANs, which is incremental as it builds on existing self-supervised and semi-supervised methods.
The paper tackles the problem of improving conditional GANs in semi-supervised settings by introducing a self-supervised pretext task that uses label space augmentation instead of image augmentations, resulting in performance surpassing baselines with only 20% of labelled examples on benchmarks like CelebA and RaFD.
We present a novel self-supervised learning approach for conditional generative adversarial networks (GANs) under a semi-supervised setting. Unlike prior self-supervised approaches which often involve geometric augmentations on the image space such as predicting rotation angles, our pretext task leverages the label space. We perform augmentation by randomly sampling sensible labels from the label space of the few labelled examples available and assigning them as target labels to the abundant unlabelled examples from the same distribution as that of the labelled ones. The images are then translated and grouped into positive and negative pairs by their target labels, acting as training examples for our pretext task which involves optimising an auxiliary match loss on the discriminator's side. We tested our method on two challenging benchmarks, CelebA and RaFD, and evaluated the results using standard metrics including Fréchet Inception Distance, Inception Score, and Attribute Classification Rate. Extensive empirical evaluation demonstrates the effectiveness of our proposed method over competitive baselines and existing arts. In particular, our method surpasses the baseline with only 20% of the labelled examples used to train the baseline.