what is the influence of "reduce_zero_label" parameter?

[MengHao666]

I have 3 classes in the image to classify, i.e., background, right hand, and left hand, and the labels of them in semantic maps are 0,1,2 respectively.

I am very confused about the parameter of "reduce_zero_label", as I saw the explanation in here
reduce_zero_label (bool): Whether reduce all label value by 1. Usually used for datasets where 0 is background label. Default: False.

Would it boost performance if I set reduce_zero_label as True? What sepecific influence does it make for the training and evaluation process?

[NielsRogge]

In some datasets (like ADE20k), the 0 index is used in the annotated segmentation maps for background. However, ADE20k doesn't include the "background" class in its 150 labels. Therefore, reduce_zero_label is used to reduce all labels by 1, and to make sure no loss is computed for the background class (i.e. it replaces 0 in the annotated maps by 255, which is the ignore_index of the loss function in the mmseg library).

However, other datasets use the 0 index as background class and include this class as part of all labels. In that case, reduce_zero_label should be set to False, as loss should also be computed for the background class.