Semantic segmentation dataset prepare

Hello everyone,
In semantic segmentation how I can assign the pixel a class in the label or segmented images.
As during training I flatten outputs of my model to give 21 classes(Pascal voc2012) but can’t able to figure out to change the pixel values in the segmented images according to the class …to be used later in calculating loss.
Please help.

Hello,

I think this thread can help you.

1 Like