Improving sample diversity of a pre-trained, class-conditional GAN by changing its class embeddings

Open in new window