ASC-Net: Adversarial-Based Selective Network for Unsupervised Anomaly Segmentation

We introduce a neural network framework, utilizing adversarial learning to partition an image into two cuts, with one cut falling reference distribution provided by the user. This concept tackles task of unsupervised anomaly segmentation, which has attracted increasing attention in recent years due their broad applications tasks unlabelled data. Adversarial-based Selective Cutting (ASC-Net) bridges domains cluster-based deep methods and adversarial-based anomaly/novelty detection algorithms. evaluate this model on BraTS brain tumor LiTS liver lesion MS-SEG2015 segmentation tasks. Compared existing like AnoGAN family, our demonstrates tremendous performance gains Although there is still room further improve compared supervised algorithms, promising experimental results shed light building algorithm using user-defined knowledge.