Super-resolution.
Put images to be used for training in the directory set_e.Settings.image_dir_load.
Execute make_d.py
to cut out images randomly.
(If you want to check the training status, put the images in the directory set_e.Settings.image_dir_test, and the output for each set sample interval will be saved in the directory set_e.Settings.image_dir_proc.)
Run train.py
to train.
The weights are saved in the directory set_e.Settings.weight_dir_save.
(Settings.image_dir_demo directory if you want to check the learned weights.)
When the training is complete, set the variable (generator_weight_path) to the name of the file where the weights are saved in and run sr_adv.py
.
file | description |
---|---|
make_d.py |
Make croped images |
train_set.py |
Dataset |
set_e.py |
Settings (directory) |
train.py |
Option (training) / Training |
sr_adv.py |
Generate images using trained weight |
model.py |
GAN model |
model_deepspeed.py |
GAN model (use deepspeed) |
Originally, we wanted to show an example of the output with an image of an adventure game, but due to copyright issues...