This repository provides code and pretrained models for training data-efficient Generative Adversarial Networks (GANs). The approach leverages cross-domain consistency to significantly reduce the number of labeled images required for training high-quality GANs, improving sample efficiency while maintaining generation quality.
Key Features
Efficient GAN training with fewer labeled examples
Cross-domain consistency regularization technique
Pretrained models for various datasets
Codebase implemented with detailed training scripts
Support for domain adaptation and transfer learning in GANs
Android DevHub
.png)
.png)
