Open f6ra07nk14 opened 3 years ago
When I trace the training code in model.py , I saw that "heatmap" is multiplied by 255 on the 38 line. Could this value be an arbitrary scale value, or have other physical meanings?
Thanks in advance.
No it doesn't have special meanings. Just to scale the loss values. I think there would be marginal performance difference if you do not multiply 255.
When I trace the training code in model.py , I saw that "heatmap" is multiplied by 255 on the 38 line. Could this value be an arbitrary scale value, or have other physical meanings?
Thanks in advance.