Hi! Does anyone use this transform in fastai2. I didn’t find it in 09_vision.augment.ipynb
fastai2 notebook and I am not quite sure how to implement it. Any help would be great
It’s not implemented yet, so if anyone would like to make a port to V2 with a PR, it would be most welcome.
Is this still an option? Also, was wondering if autoaugment would be a good addition to the library?
Thanks
Hi @arora_aman,
I implemented a rgb_transform on GPU, you can find the code here
What this transform does is to change only one channel values of the whole batch.
Is didn’t suggest a PR of it because I would like to implemented another version that only affects part of the batch (or at least at CPU level) but I am kind of busy these days.