3 Replies Latest reply on Jul 3, 2018 6:38 AM by Intel Corporation

    Questions about Rescaling images

    deeplearner

      Hi,

       

      I read part 14 which talks about image preprocessing.

       

      In the resizing images part, the article says - the possible reasons for resizing are to avoid possible overflow, optimization, stability issues, and so on.

       

      What do you mean by

       

      1. Are normalization techniques(samplewise centering, samplewise std normalization, featurewise centering, featurewise std. normalization)
        1. Alternative approaches to Rescaling?
        2. Or can be applied to images along with Rescaling?
        3. If we can apply along with Rescaling. Is there any hierarchy of application?
      2. possible overflow
      3. stability issues
      4. Also, you mentioned, "and so on.". Do you have any references where we can study about the possible reason and background information about this?
      5. In the section - "Samplewise Centering" -
        1. It was mentioned "Following the optimization stability considerations (get rid of vanishing or saturating values problems)" Does this mean -  ReScaling?

       

      Thanks so much for the article.