Hi,
The dataset was was made available with just 4 channels.
Image.open() will works because each .tif has just 1 band. My other story (https://towardsdatascience.com/how-to-create-a-datablock-for-multispectral-satellite-image-segmentation-with-the-fastai-v2-bc5e82f4eb5) shows another way to deal with n-channels, where I open the numpy version of the image.
The augmentations work for 10 channels also.
