In this project, I trained a DCGAN using the Large-scale CelebFaces Attributes (CelebA) Dataset. The goal is to get a generator network to generate new images of faces that look as realistic as possible.
- Pre-processed Data
Here the images have already been cropped to remove parts of the images that don't include a face, then resized down to 64x64x3 NumpPy images. Some sample data is shown below.
2.Define the Model
A GAN is comprised of two adversarial networks, a discriminator and a generator.
-
Discriminator: This is a convolutional classifier without any maxpooling layers. The inputs to the discriminator are 32x32x3 tensor images. The output is a single value that will indicate whether a given image is real or fake.
-
Generator: The generator should upsample an input and generate a new image of the same size as the training data 32x32x3. This should be mostly transpose convolutional layers with normalization applied to the outputs.
3.Training
Training will involve alternating between training the discriminator and the generator.
For the discriminator, the total loss is the sum of the losses for real and fake images, d_loss = d_real_loss + d_fake_loss. The generator loss will look similar only with flipped labels. The generator's goal is to get the discriminator to think its generated images are real.
4.Generate samples from training
Training will involve alternating between training the discriminator and the generator.
The details for each step can be found in this jupyter notebook: Jupyter notebook