https://github.com/revsic/tf-alae
(Unofficial) Tensorflow implementation of Adversarial Latent Autoencoder (ALAE, Pidhorskyi et al., 2020)
https://github.com/revsic/tf-alae
Last synced: 14 days ago
JSON representation
(Unofficial) Tensorflow implementation of Adversarial Latent Autoencoder (ALAE, Pidhorskyi et al., 2020)
- Host: GitHub
- URL: https://github.com/revsic/tf-alae
- Owner: revsic
- License: mit
- Created: 2020-05-03T08:14:30.000Z (about 5 years ago)
- Default Branch: master
- Last Pushed: 2020-09-08T05:40:49.000Z (over 4 years ago)
- Last Synced: 2025-05-05T17:33:05.007Z (14 days ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 878 KB
- Stars: 10
- Watchers: 4
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# tf-alae
(Unofficial) Tensorflow implementation of Adversarial Latent Autoencoder (ALAE, Pidhorskyi et al., 2020)
- Original paper, Adversarial Latent Autoencoder [[arXiv:2004.04467](https://arxiv.org/abs/2004.04467)]
- Full code is based on original repository [[GIT](https://github.com/podgorskiy/ALAE)].
## Usage
To train the mnist model
```bash
# to train mnist-alae
python mnist_mlp.py train
# to train style-alae
python mnist_style.py train
```
To open tensorboard summary
```
tensorboard --logdir summary
```
Currently, lsunbed-StyleALAE is experimental.
```bash
# to train lsunbed
python lsunbed_style.py train
```
To use released checkpoints, download files from [release](https://github.com/revsic/tf-alae/releases) and unzip it.
Following is example of [MNIST-MLP](https://github.com/revsic/tf-alae/releases/tag/MnistMlp).
```py
import json
from mnist_mlp import MnistAlae
with open('settings.json') as f:
settings = json.load(f)
alae = MnistAlae(settings)
alae.load_weights('./mnist_mlp/mnist_mlp')
```
## Jupyter notebook
- [mnist_expr.ipynb](./experiments/mnist_expr.ipynb): MNIST interpolation with Mlp-ALAE
- [mnist_style.ipynb](./experiments/mnist_style.ipynb): MNIST interpolation with Style-ALAE
## Learning Curve
Mlp-ALAE + MNIST
Style-ALAE + MNIST
- In the original paper, they claim that latent reconstruction has better perceptual quality than image one so that they do not use image reconstruction loss in the original repository.
- But for more easy training, this repository use image reconstruction loss as pretraining at half of the epochs in each resolution level.
## Sample
MNIST-MLP 0 ~ 4 polymorph
MNIST-Style 0 ~ 4 polymorph
