Do images with new WBs need to be normalised before being fed into a model for image classification?
a2108870 opened this issue · comments
In the paper you mention that you tested images from the imagenet with the new WB settings, do these images need to be normalised before being fed into the vgg or resnet models?
What do you mean by normalized? Any training preprocessing should be applied in testing too, that includes mean subtraction and division by std.
Thank you very much！
You are welcome!