SOFTWARE & SYSTEMS
VAE-CELEBA
A Variational Autoencoder in PyTorch trained on CelebA faces and an anime-faces dataset, with 64x64 generation examples.
documented
- Year
- 2020
- Framework
- PyTorch
GENERATED CELEBA FACES (100 EPOCHS)
VARIATIONAL AUTOENCODER FOR FACE GENERATION
01 / The build
Behind the project.
Implementation of a Variational Autoencoder (VAE) in PyTorch. Trained on the CelebA celebrity-faces dataset and on an anime-faces dataset, generating 64x64 samples. The repository includes the model definition (VAe.py), a dataset loader (dataset.py), and example outputs after 100 epochs on CelebA and 60 epochs on the anime dataset.
02 / In the workshop
Project media.
3 photos & media · Tap to expand
Technical details
Under the hood.
- Framework
- PyTorch
- Model
- Variational Autoencoder (VAE)
- Datasets
- CelebA, anime faces
- Output
- 64x64 generated samples
Notes & outcomes
What came out of it.
- Generated CelebA faces after 100 epochs of training.
- Generated anime faces after 60 epochs of training.
- Separate dataset loader module (dataset.py).
Keep exploring