How do you get the pairwise constraints?

Thank you very much for your work, but I have some questions！
I noticed that in line"330" of "Train.py": labelInfoListSimU, labelInfoListDissimU = createLabelInfoList(imgBatch3, labBatch3)
. It seems that you have used the label information in "labBatch3". But you have written in the PDF files that you got the pairwise constraints by the predicted label information.
Can you answer my question？