Method and apparatus with image segmentation

A processor-implemented learning method for an image segmentation, the learning method comprising:

• training a discriminator neural network configured to discriminate between a feature extracted from a training image and a feature extracted from a target image;
• training, based on a loss of the discriminator neural network, first duplicate layers, as duplications of trained first layers, so that a second feature extracted from the target image by the trained first duplicate layers is matched to a first feature extracted from the training image by the trained first layers, the trained first layers being included in a first neural network and configured to extract the first feature from the training image;
• regularizing the trained first duplicate layers so that a similarity between the first feature and a third feature extracted from the training image by the regularized first duplicate layers meets a threshold;
• and
• training second duplicate layers, as duplications of trained second layers, to correspond to the regularized first duplicate layers, the trained second layers being included in the first neural network and configured to segment the training image based on the first feature.

A processor-implemented learning method for an image segmentation, the learning method comprising:

• extracting, by duplicate feature extraction layers, second features from target images and third features from the training images;
• segmenting, by duplicate segmentation layers, the target images based on the second features and the training images based on the third features;
• training the duplicate feature extraction layers at least until one of the second features matches one of first features extracted from training images by initial feature extraction layers;
• and
• regularizing the duplicate feature extraction layers at least until a similarity between another one of the first features and one of the third features meets a threshold.