Adjusting the Generative adversarial network example

Hi. I want to feed grayscale images to the GAN and get grayscale images as output, but the generator layer is defined in a way that produces RGB images. How can I change this layer to produce grayscale images? Also, is there any other modifications needed in the training GAN provided by MATLAB in the following link?

https://www.mathworks.com/help/deeplearning/ug/train-generative-adversarial-network.html

I have tried changing the filter numbers of the last layers of both generator and discriminater from 3 to 1, but I end up with plain blue images during the learning process.

[Eaby Babu]

Hi, Have you found a solution to train gray-scale images?

[Natalia da Silva de Sá]

I fixed the blue image problem just changing the part of the code that display the images. So instead of image(I), I used imshow(I).

Answer 1

[yanqi liu]

sir，may be use

augmentedImageDatastore(data,'ColorPreprocessing','gray2rgb');