percent of similarity between images

Hi everyone

I have a set of 8 different types of images, I need to develop a system that allows from a new image, how much look like to the differents categories, I mean as an example:

Looks like "A type image" - 70%

Looks like "B type image" - 20%

Looks like "C type image" - 8%

.. And so on

I already used bag of words and pre trained alexnet NN for classification, and both work pretty well, but it only chosed one of the different cateogories. One of the ideas, that I have, is used Fuzzy logic and trough mermership functions defined this percentages of memership for each new image, but I don`t have idea if this could work and how to do it. If there is another way to accomplish this, will be great, I really appreciate all your help, thanks to all.

Answer 1

[Stephan]

When using a neural net you could use the score output argument to get this. Look at this example there is a bar chart based on the score values for classifying objects in images:

https://de.mathworks.com/help/deeplearning/ug/classify-images-from-webcam-using-deep-learning.html#d123e8937

[Juan Pablo Guamn Bernal]

great, thanks a lot