eduzhai > Applied Sciences > Engineering >

Deep Convolutional GANs for Car Image Generation

  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: In this paper, we investigate the application of deep convolutional GANs oncar image generation. We improve upon the commonly used DCGAN architecture byimplementing Wasserstein loss to decrease mode collapse and introducing dropoutat the end of the discrimiantor to introduce stochasticity. Furthermore, weintroduce convolutional layers at the end of the generator to improveexpressiveness and smooth noise. All of these improvements upon the DCGANarchitecture comprise our proposal of the novel BoolGAN architecture, which isable to decrease the FID from 195.922 (baseline) to 165.966.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...