0

In terms of training deep learning models for different types of image-related works, such as image classification, semantic segmentation, what kind of pre-processing works need to be performed?

For instance, if I want to train a network for semantic segmentation, do I need to scale the image value (normally represented as an nd-array) to [0,1] range, or keep it as [0,255] range? Thanks.

user288609
  • 12,465
  • 26
  • 85
  • 127

1 Answers1

0

There are few things that are done but really there is no set or fix set of pre-processing that is always done.

Here are some examples:

  • Subtract the mean image,
  • Divide by the variance (less common)
  • Normalize the values
  • if working with "real" images (like an image of people) then horizontal flips
  • random crops
  • translations
Steven
  • 5,134
  • 2
  • 27
  • 38
  • Hi Steven, regarding normalize the values, is that similar to what we usually do in statistics, normalize the pixel values to [0, 1]? – user288609 Nov 29 '16 at 02:33
  • yeah by normalize I mean exactly that. The values between [0,1] whereas subtract mean and divide by variance would put you in the range [-1,1] – Steven Nov 30 '16 at 05:10