Provided weights don't seem to recreate presented results.

[jdagdelen]

Hi, I just tried using the provided weights for the face model from the README and I'm getting results that aren't performing at the same level as I would expect. Are these the correct weights?

[jdagdelen]

Ok, so I've figured out it is affected by image rescaling. When I re-run with the dimensions set to 80x112 I get the following.

[remya-ravi]

can u please explain how did u run this project. did u make any modification to the code ? whicha are the dataset u used ?
please

[jdagdelen]

I adapted the code in sketchback.py and changed the input parameters to fit the appropriate image class. The data was the CUHK face data set.

[remya-ravi]

Thank you for your replay. But i need to know more details like : what do u mean by input parameters ? is it the name of image ? the function train_faces is not calling anywhere in the code. but when i call it, i got errors. and i need to know does the loss functions work . And when i change the dimension of image in the code , it does not work properly for any other image than in the cuhk dataset. and 80x112 produce an output with very low quality.