This is pytorch implementation of paper "Glow: Generative Flow with Invertible 1x1 Convolutions". Most modules are adapted from the offical TensorFlow version openai/glow.
train.py <hparams> <dataset> <dataset_root>
z_deltaand manipulate attributes with
infer_celeba.py <hparams> <dataset_root> <z_dir>
Currently, I trained model for 45,000 batches with
hparams/celeba.json using CelebA dataset. In short, I trained with follwing parameters
|image_shape||(64, 64, 3)|
|flow_permutation||invertible 1x1 conv|
|batch_size||12 on each GPU, with 4 GPUs|
Following are some samples at training phase. Row 1: reconstructed, Row 2: original.
Use the method decribed in paper to calculate
z_neg for a given attribute.
z_delta = z_pos - z_neg is the direction to manipulate the original image.
Smiling (from negative to positive):
Young (from negative to positive):
Pale_Skin (from negative to positive):
Male (from negative to positive):
There might be some errors in my codes. Please help me to figure out.