Variational Autoencoders (VAEs) are generative models which combine concepts from both deep learning and statistical inference. They can be used to learn a low dimensional representation of complex, high dimensional data. One popular application is image generation. In this project I implemented a VAE that can learn the distribution of pixels in images, playing with different datasets and distributions. The project was developed during an internship at the Romanian Institute of Science and Technology.
The project is implemented in Python using TensorFlow.
For details on how to run the code and details about the performed experiments, see documentation.pdf.
- Tutorial on Variational Autoencoders
- Introduction to Variational Autoencoders
- Auto-Encoding Variational Bayes
- Tutorial - What is a variational autoencoder? and github link
- Variational Autoencoder: Intuition and Implementation
- Introducing Variational Autoencoders (in Prose and Code)
- Variational Autoencoder in TensorFlow