We propose a method for semi-supervised training of structured-output neural networks.
Inspired by the framework of Generative Adversarial Networks (GAN), we train a discriminator network to capture the notion of `quality' of network output.
To this end, we leverage the qualitative difference between outputs obtained on the labelled training data and unannotated data.
We then use the discriminator as a source of error signal for unlabelled data.
This effectively boosts the performance of a network on a held out test set.
Initial experiments in image segmentation demonstrate that the proposed framework enables labelling two times less data than in a fully supervised scenario, while achieving the same network performance.