Generating Avatars from Real Life Pictures
We strive to transfer realistic features from photos to avatar styles, like Bitmojis, or Facebook Avatars, enhancing their customization.
generative-adversarial-networks computer-vision avatars image-to-image-translation autoencoders wandb done library research code demo


In recent years, methods for image generation have exploded giving birth to many applications. At the same time, digital avatars have become increasingly popular in social networks. Thus, we see a big opportunity to use generative networks to improve and automatize avatar creation.


To explore methods like GANs and VAEs to create digital avatars, being able to specify their general style as well as their specific features. This will make them more similar to the people they are intended to resemble, and could even automatize their creation from real life pictures.

Datasets to use: CelebA, Google facial expression comparison dataset, YouTube Faces Dataset with Facial Keypoints.

Future Possibilities

Since both the software techniques and avatar creation are relatively new topics, this work can then be improved and extended to the many applications into which these topics will grow.

Reference paper:

Don't forget to tag @Leinadh , @StevRamos , @davfre98 , @weirdfish23 in your comment, otherwise they may not be notified.

Authors original post
Computer Science Student | Data & Analytics Intern | Looking forward to improving my deep learning skills and developing interesting applications
Computer Science student - Data Scientist Intern
Electrical engineering student | Interested in all things AI, from hardware architectures to high-level algorithms | Looking forward to apply deep learning techniques to real world problems.
Share this project
Similar projects
PyTorch Tutorial for Deep Learning Researchers
This repository provides tutorial code for deep learning researchers to learn PyTorch.
A Tensorflow-based framework to ease the training of generative models
Tensorflow 2.0 Implementation of the paper Image-to-Image Translation using Conditional GANs by Philip Isola, Jun-Yan Zhu, Tinghui Zhou and Alexei A. ...
Adversarial Latent Autoencoders
Introducing the Adversarial Latent Autoencoder (ALAE), a general architecture that can leverage recent improvements on GAN training procedures.
Top collections