Question regarding code

[D0miH]

Hello, first of all thanks for publishing this awesome project @apchenstu.

I was looking through the code of the facialDetails.py file and have a question regarding this python script. During creation of the patches you stitch them together using the weights and accumulate the weights in the count array. But why are you dividing each pixel by the weights in this two lines?: https://github.com/apchenstu/Facial_Details_Synthesis/blob/def9bfe044790d771d15dc4a62bb47b78d7d6153/src/facialDetails.py#L190-L191

My guess is that the result is somehow "normalized" but how exactly is it normalized? Such that it is in a range between 0 and 255?

[apchenstu]

Hi, our method is to predict facial detail path-ly (256*256) from a given image, so there is some overlap between paths. we use a "count" variable to accumulate the overlap for each pixel and dividing it at the final result.