-2

I'm trying to simulate this article.

In order to make images noisy and based on the article I need to add manually deterministic distribution noise to Mnist dataset. The article says

" This noise has been added manually using deterministic distribution"

in page 973 at the first row. I looked for it almost in everywhere, but was unable to find how to do it. They usually use this distribution for other goals, not to make images noisy.

This article adds and measures noise by percent, for example a 50% noisy image.

How can we add it with percent in python? I really need help with that.

morgan121
  • 2,213
  • 1
  • 15
  • 33
mahyar sadeghi
  • 77
  • 2
  • 8
  • 1
    What on earth do you (or the authors) mean by "add manually deterministic distribution"? Do you realize that deterministic distribution is a distribution yielding a fixed value with probability 1, i.e., it means a constant? So you're saying you want to manually add a constant to your data? – pjs Feb 05 '19 at 21:01
  • Actually I'm good at distributions...but I'm looking for a code to make my data noisy and as my friend said it is not that clear. – mahyar sadeghi Feb 05 '19 at 21:16
  • 1
    Whether you're good at distributions or not, you need to clarify what you're actually asking for. – pjs Feb 05 '19 at 21:30
  • Anyway...do u have any advise?how can i make this noise? – mahyar sadeghi Feb 05 '19 at 21:34
  • 1
    I can't give advice if I can't understand what you're asking. Once again, I'm asking you to clarify. If you don't know, your best bet is to contact the original authors and ask them what they meant, since they're the ones who used unclear terminology without any explanation. – pjs Feb 05 '19 at 21:58
  • Ok thanks..i will do these – mahyar sadeghi Feb 05 '19 at 22:08

1 Answers1

0

They never really make it clear, but it looks like they're randomly assigning pixels a value of 0 for each of their inputs. In the case you mention, 50% of the pixels were assigned 0, and those pixels were chosen at random.

Jordan Singer
  • 567
  • 5
  • 11
  • Is it acceptable to add other types of noise to this dataset and use it?after simulating this article we wanna write a new article based on this.but I'm worried about it that maybe journals reject our article because our noise types are different – mahyar sadeghi Feb 05 '19 at 21:40
  • What do you mean by "acceptable"? There's nothing wrong with using different types of noise. Gaussian? Perlin? Go to town. – Jordan Singer Feb 05 '19 at 22:03
  • I know...but my first step is simulating this artricle..if i use other noises,maybe i get different results from article...in this case my data and models are not validated – mahyar sadeghi Feb 05 '19 at 22:07
  • I'm not sure what you're asking. If you want to do what they did, then you can just do that, yeah? – Jordan Singer Feb 05 '19 at 22:11
  • If you want to write a new paper,first step is simulating other articles and get the exact results.And after that u can add your idea to that simulated results to see how it can be changed...right? My study is about denoising techniques to get good image classification accuracy...I decided to work on this article and then improve it to get better results.I could simulate models in it but now my problem is the noise type...i don't understand how they added in images.but i think my only chance is send an email to real authors because as you said it is really unclear.anyway thanks for your help – mahyar sadeghi Feb 05 '19 at 22:52