The Generative Adversarial Network (GAN) consists of a generator and a discriminator which are made up of small residual network architecture. It started giving out quite good results after training for only 20 epochs. The dataset can be found here: https://www.kaggle.com/paramaggarwal/fashion-product-images-small. It consists of smaller resolution images of the original dataset. The GAN might produce better results if trained on high resolution images. The generator and discriminator models were trained at tandem with a fixed learning rate: 0.0002.
sudipto153 / dress_gan Goto Github PK
View Code? Open in Web Editor NEWThe Generative Adversarial Network consists of a generator and a discriminator which are made up of small residual network architecture. It started giving out quite good results after training for only 20 epochs.