For once, when doing the semantic segmentation, image pixel information, than those obtained by the label is not 0 or 1, is a decimal, isn't there any way to process the output forecast figure directly to 0 or 1.