Overview
• GAN have two characters contains generator and discriminator.
- Generator generate images from latent code, Discriminator classify images
into categories.
- The primary goal of Generator is fool the discriminator, make loss of discriminator maximum.
- In contrast, the main goal of discriminator is correctly classify whether a image(or data) is real(from original dataset) or fake(made by generator).
Dataset
• Dataset: CelebA Face Dataset
• CelebFaces Attributes Dataset (CelebA) is a large-scale face attributes dataset
with more than 200K celebrity images, each with 40 attribute annotations.
• The images in this dataset cover large pose variations and background clutter. CelebA has large diversities, large quantities, and rich annotations, including
• 10,177 number of identities,
• 202,599 number of face images, and
• 5 landmark locations, 40 binary attributes annotations per image. • Original Size: 218×178
Your task
- Skeleton Code: https://colab.research.google.com/drive/1mOjpQEfI2ivYtHndNytknfK nclFJ-olE#scrollTo=aGCZQxZSfONu
- Implement a basic DCGAN
- Improve performance of DCGAN
• Use SELU as activation function
• Adopt training process of Relativistic GAN - More advanced Modifications
- Read notebook to get more details
Things you cannot do
- You cannot submit results predicted by others.
- You cannot copy trained models from others.
- You cannot copy code from others, internet, GitHub …
- You cannot collect more images to train your model in order to boost performance.
Any violation will result in 0 score!
