This repository is a training exercise in exploring various techniques for image generation, such as GANs, autoencoders, and diffusion models. The goal is to experiment with these methods and deploy the results on a web interface that works seamlessly without a server. The models used are lightweight to ensure fast loading and responsiveness.
- Current Models: Generative Adversarial Networks (GANs)
- Current Datasets: MNIST, Fashion MNIST
- Web Deployment: The generated models are deployed on a static webpage that can be accessed here.
- Additional Models:
- Autoencoder
- Diffusion Model
- StyleGAN or similar architecture
- Additional Datasets:
- CIFAR-10
This project is a learning exercise focused on generative models and their deployment. It also serves as preparation for a research project that will start in February, aimed at predicting and analyzing facial expressions using diffusion models.
- Models are pre-trained and optimized for lightweight deployment.
- No server is required; the entire project runs in-browser, making it highly accessible and easy to maintain.
Visit the web interface to explore the current implementations. More models and datasets will be added over time.