Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
---
|
4 |
+
|
5 |
+
# User-Controllable Latent Transformer for StyleGAN Image Layout Editing
|
6 |
+
|
7 |
+
Yuki Endo: "User-Controllable Latent Transformer for StyleGAN Image Layout Editing," Computer Graphpics Forum (Pacific Graphics 2022) [[Project](http://www.cgg.cs.tsukuba.ac.jp/~endo/projects/UserControllableLT)] [[PDF (preprint)](http://arxiv.org/abs/2208.12408)]
|
8 |
+
|
9 |
+
## Prerequisites
|
10 |
+
1. Python 3.8
|
11 |
+
2. PyTorch 1.9.0
|
12 |
+
3. Flask
|
13 |
+
4. Others (see env.yml)
|
14 |
+
|
15 |
+
## Preparation
|
16 |
+
Download and decompress <a href="https://drive.google.com/file/d/1lBL_J-uROvqZ0BYu9gmEcMCNyaPo9cBY/view?usp=sharing">our pre-trained models</a>.
|
17 |
+
|
18 |
+
## Inference with our pre-trained models
|
19 |
+
We provide an interactive interface based on Flask. This interface can be locally launched with
|
20 |
+
```
|
21 |
+
python interface/flask_app.py --checkpoint_path=pretrained_models/latent_transformer/cat.pt
|
22 |
+
```
|
23 |
+
The interface can be accessed via http://localhost:8000/.
|
24 |
+
|
25 |
+
## Training
|
26 |
+
The latent transformer can be trained with
|
27 |
+
```
|
28 |
+
python scripts/train.py --exp_dir=results --stylegan_weights=pretrained_models/stylegan2-cat-config-f.pt
|
29 |
+
```
|
30 |
+
To perform training with your dataset, you need first to train StyleGAN2 on your dataset using [rosinality's code](https://github.com/rosinality/stylegan2-pytorch) and then run the above script with specifying the trained weights.
|
31 |
+
|
32 |
+
## Citation
|
33 |
+
Please cite our paper if you find the code useful:
|
34 |
+
```
|
35 |
+
@Article{endoPG2022,
|
36 |
+
Title = {User-Controllable Latent Transformer for StyleGAN Image Layout Editing},
|
37 |
+
Author = {Yuki Endo},
|
38 |
+
Journal = {Computer Graphics Forum},
|
39 |
+
volume = {41},
|
40 |
+
number = {7},
|
41 |
+
pages = {395-406},
|
42 |
+
doi = {10.1111/cgf.14686},
|
43 |
+
Year = {2022}
|
44 |
+
}
|
45 |
+
```
|
46 |
+
|
47 |
+
## Acknowledgements
|
48 |
+
This code heavily borrows from the [pixel2style2pixel](https://github.com/eladrich/pixel2style2pixel) and [expansion](https://github.com/gengshan-y/expansion) repositories.
|