The test data normalization problem in Block-level JND Predictive Model/S4_Test.py. I see that the 118th line of code uses the mean and variance of ‘X_Train’ for normalization, and the TPI information is not normalized. You are in Model.py Normalize the test and validation data using the mean and variance of the training set. Shouldn't they be performed independently? This is inconsistent with the data processing methods I usually see. Can you explain it? Finally, I would like to ask, I used my own video test and used the mean and variance of my own video for normalization. I found that the results of JND1, JND2, and JND3 were a bit strange. I am not sure whether it is because of the normalization used. The mean and variance are inconsistent with what you used during training. I hope you can explain it, thank you.