I made the codes to aim at synthesizing face images
- PyTorch
- Python
- tqdm
- pillow
- matplotlib
- numpy
- 1 gpu (GeForce RTX 3080 ti)
- batch size is 1
To train quickly, only 10,000 images were used.
- Image : CelebAMask-HQ
- Segmentation map : Merging masks and colorizing them
Something to keep in mind : It requires only png
image. convert jpg image to png image.py
python main.py
python test.py
- pretrained
- segmap_label.txt : You can use the same file used for train.