albumentations-team / albumentations

Fast and flexible image augmentation library. Paper about the library: https://www.mdpi.com/2078-2489/11/2/125
https://albumentations.ai
MIT License
14.31k stars 1.65k forks source link

Albumentations for black and white drawing images like Quickdraw and Fashion MNIST datasets #966

Closed cgnorthcutt closed 3 years ago

cgnorthcutt commented 3 years ago

Thank you for this great package. I'm trying to use Albumentations for augmenting black and white drawing images like Quickdraw and Fashion MNIST datasets, but most of the examples seem focused on color images. What do you recommend?

ternaus commented 3 years ago

I would say that the majority, if not all of the augmentations will work for grayscale images if your treat them as RGB with 3 channels obtained from the grayscale image, say with rgb_image = cv2.cvtColor(grayscale_image, cv2.COLOR_GRAY2RRGB)

You may also check machine learning competitions at Kaggle where participants worked with the medical or other grayscale images.

P.S. I would guess, that you may get more detailed advice at Stackoverflow.