peterzhousz / bdinvert_release

BDInvert: GAN Inversion for Out-of-Range Images with Geometric Transformations

This repository contains the accompanying code for GAN Inversion for Out-of-Range Images with Geometric Transformations, ICCV 2021

Prerequisites

• Ubuntu 18.04 or higher
• CUDA 10.0 or higher
• pytorch 1.6 or higher
• python 3.7 or higher

Installation

pip install -r requirements.txt

Usage

Train

1. Change directory into BDInvert.

cd BDInvert

2. Train base code encoder.

python train_basecode_encoder.py

• --model_name : You can change backbone StyleGAN model, which located at model zoo.
• --basecode_spatial_size : You can change spatial resolution of basecode.

3. Find pnorm parameters.

python pca_p_space.py

Download pretrained base code encoder.

Download and unzip under BDInvert/pretrained_models/.

Encoder Pretrained Models	Basc Code Spatial Size
StyleGAN2 pretrained on FFHQ 1024, 16x16	16x16

Test

• Default test setting use StyleGAN2 pretrained on FFHQ1024 and use basecode spatial size as 16x16.

1. Change directory into BDInvert.

cd BDInvert

2. Make image list.

python make_list.py --image_folder ./test_img

3. Embed images into StyleGAN's latent codes.

python invert.py --encoder_pt_path {encoder_pt_path}

• --image_list : Inversion target image list generated from above step 2. Default is ./test_img/test.list
• --weight_pnorm_term : As recently well known, there is a trade-off between editing quality and reconstruction quality. This argument controls this trade-off.

4. Edit embedded results.

python edit.py {inversion directory}

• --edit_direction : You can change edit direction, which located at BDInvert/editings/interfacegan_directions

Small changes

• We changed the detail code regularization method from hard clipping in P-norm+ space to L2 norm regularization, following the update of the original paper.
• Due to this change, new hyperparameter, weight_pnorm_term, has been added.

License

This software is being made available under the terms in the LICENSE file.

Any exemptions to these terms requires a license from the Pohang University of Science and Technology.

Useful Links

• Author Homepage
• POSTECH CG Lab.

Acknowledgments

NOTE

• Our implementation is heavily based on the "GenForce Library".
• Interface GAN editing vectors are from "encoder4editing".