Fashion MNIST dataset is directly downloaded from tf.keras.datasets
Composed of 60,000 training images and 10,000 test images
Dimension of each image is 28x28
Random Gaussian noise is added to make sure we get realistic images and not the exact images
To improve DCGAN model performance (use GPU/Colab for these cases)
- Increase the number of epochs
- Increase the number of layers for Generator and Discriminator
To use custom images of different dimensions
- Make changes in input size of Discriminator
- Modify layers in both Generator and Discriminator accordingly