Are the probabilities in prior_probs.npy smoothed with a Gaussian kernel?

richzhang / colorization

Automatic colorization using deep neural networks. "Colorful Image Colorization." In ECCV, 2016.

http://richzhang.github.io/colorization/

BSD 2-Clause "Simplified" License

3.35k stars 928 forks source link

Are the probabilities in prior_probs.npy smoothed with a Gaussian kernel? #47

Closed mingo-x closed 6 years ago

mingo-x commented 6 years ago

Hi,

I didn't find the code to generate the _priorprobs.npy file, so I decided to ask here. Are the probabilities in the _priorprobs.npy file raw empirical probabilities? Or are they smoothed probabilities?

Many thanks!

richzhang commented 6 years ago

Hi, they are smoothed with the same Gaussian used to compute soft targets during training.

mingo-x commented 6 years ago

Thank you very much for the timely reply!