Skip to content

shining0611armor/Simple-Implementation-for-VAE-CVAE-and-VQVAE

Repository files navigation

🚀 Simple Implementation for VAE, CVAE, and VQVAE

Exploring the Variational Autoencoder (VAE) family

In this repository, we provide a minimal implementation of the VAE family, including VAE, CVAE, and VQVAE. These implementations are applied to both Anime-Face and Cartoon-Face datasets. Let's embark on this journey from zero to hero! 🌟

📂 Implementations

🔹 VAE

The Variational Autoencoder (VAE) is a generative model that learns a probabilistic mapping from a data space to a latent space. Below is an image generated using VAE:

VAE generated image VAE generated image

🔸 CVAE

The Conditional Variational Autoencoder (CVAE) extends VAE by conditioning the generative process on additional information. Here’s an image generated according to each label:

CVAE

🔹 VQVAE

The Vector Quantized Variational Autoencoder (VQVAE) introduces discrete latent variables to the VAE model, enabling more efficient and powerful representations. Below is an image reconstructed using VQVAE:

reconstruction with VQVAE

📫 Contact

Feel free to reach out if you have any questions or suggestions:


Happy Learning! 😊

About

Exploring the Variational Autoencoder (VAE) family

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published