Method and apparatus with image segmentation
A processor-implemented learning method for an image segmentation includes training first duplicate layers, as duplications of trained first layers of a pre-trained model, so that a second feature extracted from a target image by the trained first duplicate layers is matched to a first feature extracted from a training image by the trained first layers; 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 of the pre-trained model, to be configured to segment the target image based on the regularized first duplicate layers, the trained second layer being configured to segment the training image.
1. A processor-implemented learning method for an image segmentation, the learning method comprising:
training first duplicate layers, as duplications of trained first layers of a pre-trained model, so that a second feature extracted from a target image by the trained first duplicate layers is matched to a first feature extracted from a training image by the trained first layers;
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 of the pre-trained model, to be configured to segment the target image based on the regularized first duplicate layers, the trained second layer being configured to segment the training image.
2. The method of claim 1 , further comprising:
generating a new model by combining the regularized first duplicate layers and the trained second duplicate layers; and
segmenting the target image by semantic segmentation using the generated new model.
3. The method of claim 1 , wherein the training image is an image of a first domain environment different from a second domain environment of the target image.
4. The method of claim 1 , wherein the training of the first duplicate layers comprises training the first duplicate layers based on a loss corresponding to a difference between the first feature and the second feature.
5. The method of claim 4 , wherein the training of the first duplicate layers comprises training a discriminator neural network to discriminate between the first feature and the second feature.
6. The method of claim 1 , wherein:
the trained second duplicate layers are configured to segment the training image based on the third feature; and
the training of the second duplicate layers comprises training the second duplicate layers so that a label of the segmented training image is matched to a predefined label corresponding to the training image.
7. The method of claim 1 , wherein the training of the second duplicate layers comprises training the second duplicate layers to be configured to segment the target image based on another second feature extracted from the target image by the regularized first duplicate layers.
8. The method of claim 1 , further comprising:
identifying from a pre-trained neural network, as the pre-trained model, the trained first layers and the trained second layers based on the training image and predefined labels corresponding to the training image.
9. The method of claim 1 , wherein the second duplicate layers are configured to segment the target image by semantic segmentation based on the second feature.