Should I do the reverse-normalisation after pass the last activation function

Question

I have been working on frame interpolation on Pytorch, and I have one question to ask about image normalization. Before train a model, I normalized the image dataset, with mean=[0.3852, 0.3699, 0.3618] and std=[0.2444, 0.2472, 0.2391] (this is the mean and std of my whole dataset) When training model, I passed feature maps through a sigmoid function. Now, I want to visualize the last output. Should I do the reverse-normalization before visualization? The same question when the last activation is tanh function.