-2

In my work, I have to do image matching, provided one is real image and other is from different source, but the the size of images can be different, both images can be taken from different angles(very small change in angle) and other image can have some extra smiley on it.

Note: Images contain texts such as product name and description, and I have only one photo per product(total 500 products).

How to compare both images and return the similarity?

Satyam
  • 276
  • 3
  • 5
  • Stack Overflow is not a "make-my-code-for-me" website. People will gladly help you if you provide a piece of code that is not working. – Martial Jun 15 '21 at 13:39
  • thanks @Martial but, I didn't ask for code anywhere, I just asked for approach if you can see. – Satyam Jun 15 '21 at 13:58

1 Answers1

0

You can use Simple Approach "Structural Similarity Index". Or use Deep Learning Approach you can use "Siemese Network". This method is used for image similarity. If you have no time to write a code and implement it, you can search in github and ready to use.

skidipap
  • 1
  • 1
  • I have already tried SSI and transfer learning approaches to do it but they are not yielding good accuracy as two product can have similar images but can differ only by name., for Siemese I do not have enough data to train. – Satyam Jun 17 '21 at 07:02