numberGAN

First implementation of a GAN architecture to generate new numbers from the MNIST dataset

Here's a gif of the numbers created by the GAN

The numbers are fairly visible at the moment. With more training, I will be able to get more realistic looking numbers generated.

If you want to test the output of the trained GAN call the generator below and it will produce a random number

    noise = tf.random.normal([1, 100])
    generated_image = generator(noise, training=False)
    plt.imshow(generated_image[0, :, :, 0], cmap='gray')