The goal of this project is to build a GAN that is capable of creating car images in a realistic way. To do so, the project will follow a WGAN architecture and use the data coming from The Car Connection webpage.
The dataset has more than 60,000 images, that will be classified previously into different categories attending to the point of view where the car images has been taken (front, rear, left, etc.). Once the images has been classified, the WGAN architecture uses them in order to generate the new images
The main objectives that I have set for this project are using TensorFlow >= 2 and use the aforementioned WGAN architecture. As an extra objective, I would like to implement the BoolGAN Architecture defined in this paper.