When you scale by 255, some of the pixels which already have a value close to zero (such as black part of the image) will have a very small number as the value. However, when you go on to substract the imagenet mean, some pixels will have negative values. Is this as intended?
When you scale by 255, some of the pixels which already have a value close to zero (such as black part of the image) will have a very small number as the value. However, when you go on to substract the imagenet mean, some pixels will have negative values. Is this as intended?