This repository contains an implementation of easy_BigGAN, a simplified but feature-rich BigGAN for generating high-quality images. The implementation includes key features such as conditional batch normalization, residual blocks, truncated noise sampling, and gradient penalty.
· Conditional Batch Normalization: Incorporates class information into the normalization process.
· Residual Blocks: Enhances the network’s ability to generate high-resolution images.
· Truncated Noise Sampling: Balances the diversity and quality of generated images.
· Gradient Penalty: Improves training stability.
1.Prepare the dataset (CIFAR-10)
2.Train the model
The file BigGAN-2.py introduces attention mechanisms and Batch Normalization, and updates the upsampling mechanism.