Generative adversarial network (GAN)-based system for generating color image from edge image
A generative adversarial network (GAN)-based system for generating a color image from an edge image includes: a first GAN training a model for converting the edge image into an intermediate image; and a second GAN training a model for converting the intermediate image into the color image, wherein an entropy of the intermediate image corresponds to a value between an entropy of the edge image and an entropy of the color image.
1. A generative adversarial network (GAN)-based system for generating a color image from an edge image, the GAN-based system comprising:
a first GAN training a model for converting the edge image into at least one intermediate image; and
a second GAN training a model for converting the intermediate image into the color image,
wherein an entropy of the intermediate image corresponds to a value higher than an entropy of the edge image, and the entropy of the intermediate image corresponds to a value lower than an entropy of the color image,
wherein the first GAN includes:
a first generator training a model for converting the edge image into the intermediate image; and
a first discriminator training a model for discriminating between an image generated by the first generator and a sample image representing the intermediate image,
wherein the first generator includes a first encoder including a plurality of convolution layers,
wherein the plurality of convolution layers of the first encoder include at least one dilated convolution layer, and
wherein the at least one dilated convolution layer is a layer in which a dilation rate defining an interval between kernels is introduced.
2. The GAN-based system of claim 1 , wherein the intermediate image corresponds to at least one of a gray image and a luminance component image.
3. The GAN-based system of claim 2 , wherein the luminance component image includes only data corresponding to a Y component corresponding to a value obtained by scaling a brightness of light to a range of 0 to 255 in a YUV color space.
4. The GAN-based system of claim 1 , wherein the first generator further includes:
a first decoder including a plurality of deconvolution layers,
wherein an intermediate result image generated when the edge image passes through at least one of the plurality of convolution layers in the first encoder is used as an input image of the first decoder.
5. The GAN-based system of claim 1 , wherein the second GAN includes:
a second generator training a model for converting the intermediate image into the color image; and
a second discriminator training a model for discriminating between an image generated by the second generator and a sample image representing the color image.