/DTLS

Domain Transfer in Latent Space

Primary LanguagePython

Dataset

To prepare FFHQ dataset, you can follow: FFHQ

Training

Follow the command lines below

DTLS (16 --> 128)

python main.py --mode train --hr_size 128 --lr_size 16 --stride 4 --train_steps 100001 --save_folder 'DTLS_16_128' --data_path 'your_dataset_directory' --batch_size 16

Evaluation

Follow the command lines below

DTLS 16 --> 128

python main.py --mode eval --hr_size 128 --lr_size 16 --load_path 'pretrained_weight/DTLS_128.pt' --save_folder 'DTLS_16_128_results' --input_image 'your_images_folder'

my train python main.py --mode train --hr_size 128 --lr_size 16 --stride 4 --train_steps 100001 --save_folder 'DTLS_16_128' --data_path ./training_set/ --batch_size 16

python main_smiling.py --mode train --hr_size 128 --lr_size 16 --stride 4 --train_steps 100001 --save_folder 'DTLS_smiling' --data_path ./fake_dataset_128/ --batch_size 16

python main_smiling.py --mode train --hr_size 128 --lr_size 16 --stride 4 --train_steps 100001 --save_folder 'myDTLS_smiling' --data_path ./fake_dataset_128/ --batch_size 16

python main_smiling.py --mode train --train_steps 100001 --save_folder 'myDTLS_smiling' --data_path ./fake_dataset_128/ --batch_size 16