search

lantgabor / Unsupervised-Data-Augmentation-PyTorch

Unsupervised-Data-Augmentation-PyTorch
Apache License 2.0
13 stars 1 forks source link

a question of moudle #1

Open swy0823 opened 4 years ago

swy0823 commented 4 years ago

hello,brother，i have a question thay dataset,py set from RandAugment import RandAugment but i cant fing RandAugment only find randaugment , so i set from randaugment import RandAugment, RandAugment(3,9) has error ,it need not args,so whats your meaning of RandAugment(3,9).from a rookie ,thanks for your soon answer

lantgabor commented 4 years ago

Hello, when I made this repo there was no official randaugment implementation in pytorch that was installable through pip. I used some code from other repos and the original tensorflow lib. You can try this code by replacing the RandAugment with the now available option, however, none of the pytorch implementation does agumentation the same way RandAugment paper described.

Try using some implementation and add it to compse([]). pip install randaugment gives a version which did not define N and M variable for the constructor. You may try adding them yourself if it is what you need.

dengbaowang commented 3 years ago

Hello, brother, I get the same problem about the version of randaugment. 😔

lantgabor commented 3 years ago

Hello, try to use a different implementation of randaugment and import that. I may update the repo to solve this issue. The original UDA paper has new versions as of 2020 too.