Train classifier with GAN generated images

I would like to compare a classifier trained with original images and then I would like to train the same classifier with GAN generated images + original images.

Could you please guide me on how to train the classifier with GAN generated images + original images.

Answer 1

[Muskan]

Hi Anbarasi,

As per my understanding of the question, you can followmay try the following steps for help with your problem to resolve the issue you are facing:

i) Collect and pre-process your original images dataset. Then train your original image classifier using any algorithm you wish to.

ii) Using the GAN model, generate a new dataset of images.

iii) Using the "imageDatastore" function combine the original and GAN generated images. Then, train a new classifier using the combined dataset by using the same algorithm.

iv) Now, you can evaluate the performance of both the classifiers using a test dataset.

You can refer to the following documentation for "imageDatastore" function:

Datastore for image data - MATLAB (mathworks.com)

I hope the above information helps in resolveing your query.

Thanks,

Muskan.