English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

Thanks to their ability to learn data distributions without requiring paireddata, Generative Adversarial Networks (GANs) have become an integral part ofmany computer vision methods, including those developed for medical imagesegmentation. These methods jointly train a segmentor and an adversarial maskdiscriminator, which provides a data-driven shape prior. At inference, thediscriminator is discarded, and only the segmentor is used to predict labelmaps on test images. But should we discard the discriminator? Here, we arguethat the life cycle of adversarial discriminators should not end aftertraining. On the contrary, training stable GANs produces powerful shape priorsthat we can use to correct segmentor mistakes at inference. To achieve this, wedevelop stable mask discriminators that do not overfit or catastrophicallyforget. At test time, we fine-tune the segmentor on each individual testinstance until it satisfies the learned shape prior. Our method is simple toimplement and increases model performance. Moreover, it opens new directionsfor re-using mask discriminators at inference. We release the code used for theexperiments at https://vios-s.github.io/adversarial-test-time-training.

Stop Throwing Away Discriminators! Re-using Adversaries for Test-Time Training