This repository provides an implementation of Conditional Generative Adversarial Networks (CGANs) using Keras, trained on the MNIST and CIFAR-10 datasets. CGANs allow for conditional generation of images based on class labels, enabling the model to generate images of a specified class.
This implementation:
- Trains a CGAN model on the MNIST and CIFAR-10 datasets.
- Supports flexible conditioning on class labels.
- Provides visualization of generated samples during training.
model.py- Core script containing the implementation of the Generator, Discriminator and CGAN architectures.train.py- Script to train the CGAN model on MNIST and CIFAR-10 datasets.datasets.py- Contains code for loading and processing the datasets.utils.py- Contains some utility functions.main.py- The starting point of the project.state/- Directory to save the trained CGAN model weights and progression outputs.
