search

albumentations-team / autoalbument

AutoML for image augmentation. AutoAlbument uses the Faster AutoAugment algorithm to find optimal augmentation policies. Documentation - https://albumentations.ai/docs/autoalbument/
https://albumentations.ai/docs/autoalbument/
MIT License
199 stars 20 forks source link

missing parameter of input channels #19

Closed huanghjA closed 1 year ago

huanghjA commented 3 years ago

How to change the input channel into 1 as my dataset are not RGB images.

Dipet commented 3 years ago

Just simple use single channel image. But be careful, because some transforms works only with RGB images.

huanghjA commented 3 years ago

Thanks for your reply! However, it seems that the input channel is set to 3 as default and I don't find a way to change it.

when I set my input image with shape (960, 1280, 3) as RGB, It can work without runtime error. If I change the shape into (960, 1280, 1) as a gray image, the following error occur: ValueError: non-broadcastable output operand with shape (960,1280,1) doesn't match the broadcast shape (960,1280,3)

Dipet commented 3 years ago

Describe what transforms you are using. Perhaps some of them do not support grayscale images, or maybe there is a bug in them.

huanghjA commented 3 years ago

In fact, this issue has been post in #8 . In the filesearch.yaml, I delete all transforms but keep