Awesome
WAE-pytorch
Pytorch implementation of WAE-MMD(paper).
Dependencies
python 3.6.4
pytorch 0.3.1.post2
visdom
<br>
Usage
- download
img_align_celeba.zip
andlist_eval_partition.txt
files from here, makedata
directory, put downloaded files intodata
, and then run./preprocess_celeba.sh
. for example,
.
└── data
└── img_align_celeba.zip
└── list_eval_partition.txt
- initialize visdom
python -m visdom.server
- run by scripts
sh run_celeba_wae_mmd.sh
- check training process on the visdom server
localhost:8097
<br>
Results - CelebA
training plots
train data reconstruction
test data reconstruction
random data generation via sampling z from P(z)
<p align="center"> <img src=misc/random_sample.jpg> </p>References
- Wasserstein Auto-Encoders, Tolstikhin et al, ICLR, 2018
- Code repos : official, re-implementation, both in Tensorflow