Deep Convolutional GANs for Car Image Generation
This work addresses image generation for cars, but it is incremental as it builds upon existing DCGAN methods.
The paper tackled car image generation by proposing BoolGAN, an improved deep convolutional GAN architecture, which reduced the FID score from 195.922 to 165.966 compared to a baseline.
In this paper, we investigate the application of deep convolutional GANs on car image generation. We improve upon the commonly used DCGAN architecture by implementing Wasserstein loss to decrease mode collapse and introducing dropout at the end of the discrimiantor to introduce stochasticity. Furthermore, we introduce convolutional layers at the end of the generator to improve expressiveness and smooth noise. All of these improvements upon the DCGAN architecture comprise our proposal of the novel BoolGAN architecture, which is able to decrease the FID from 195.922 (baseline) to 165.966.