How can I implement a validation dataset?

[richiprieto]

Hello thank you very much for your code, it is very useful to study it, I wanted to ask how to implement the validation process in the code since currently there is only training and testing. Thank you very much in advance

[A03ki]

Hi, @richiprieto. Thank you for using this repository and for your positive comment.

Do you expect to split the data between training and validation data, or do you expect to load files for validation?

[richiprieto]

I would like to use new files for validation

[A03ki]

Thank you for your response.

In that case, some code needs to be corrected on this repository. The files to be modified and their parts are as follows.

• your_own_dataset/train_wgangp.py
  • Add an argument that allows passing the root name of the validation dataset.
  • Add a data loader for the validation dataset.
  • Put the validation's data loader in the function train_wgangp argument.
• your_own_dataset//train_encoder_izif.py
  • It is roughly the same as the part of your_own_dataset/train_wgangp.py that needs to be modified.
• fanogan/train_wgangp.py
  • Add an argument for the validation's data loader.
  • Add the validation process at the end of the epoch.
• fanogan/train_encoder_izif.py
  • It is roughly the same as the part of fanogan/train_wgangp.py that needs to be modified.

Feel free to ask me if you have any questions. I hope it will go well!

[A03ki]

I implemented the validation process.

You can try the following commands.

git clone -b add_validation_process https://github.com/A03ki/f-AnoGAN.git
cd f-AnoGAN
python setup.py install
cd your_own_dataset

python train_wgangp.py "your_own_dataset_dir_name/train_dir_name" --valid_root "your_own_dataset_dir_name/valid_dir_name"
python train_encoder_izif.py "your_own_dataset_dir_name/train_dir_name" --valid_root "your_own_dataset_dir_name/valid_dir_name"

Please replace your_own_dataset_dir_name/valid_dir_name with a relative path of your own validation data.

[richiprieto]

Thank you very much, you are a great programmer, as I see the code, the only thing to do to have the training values is to train with the new dataset. Nothing else changes.