keras - class_mode parameter in ImageDatagenerator

Question

I am trying to segment 5 classes in Fundus images.I use a Unet which has an output of 5 channels and have created a custom rgb_to_onehot function.I want to augment the images with Imagedatagenerator and use "flow_from_directory" which has a parameter class_mode.My question is should i set it to None or 'input'?I tried 'input' but get total black images and was wondering if i chose the correct.On the other hand by choosing None i get error I use dice loss function and dice metrics.

question from:https://stackoverflow.com/questions/66058251/class-mode-parameter-in-imagedatagenerator

Answer 1

Waitting for answers