5

Is it technically possible to find out similar products online based on an given image?

Say the first link is a hoodie for women selling on Amazon. I want to do two things.

  1. Based on the image, I want to generate features of the hoodie in addition to the ones provided from the product specs(color, size). For example, a logo in the middle and large pocket in the front
  2. I want to find out the similar product(second link) based on the image(first link). Is it feasible to find similar images with at least this degree of matching(the more the better) from a list of images? I will assume it's a failed match if finds me a women's hoodie with a big pocket in front but no logo.

https://www.amazon.com/GAP-Girls-Fleece-Pullover-Hoodie/dp/B075YPTLTQ/ref=sr_1_1_sspa?keywords=gap&qid=1552274273&s=gateway&sr=8-1-spons&psc=1

https://www.amazon.com/dp/B07F65W746/ref=twister_B07F62RWZL?th=1

user2979630
  • 201
  • 1
  • 3

1 Answers1

2

Yes, it is technically possibly. The easiest algorithm you can try out is siamese networks. Basically siamese networks try to learn a similarity measurement of two inputs (images, sounds, whatever) and work quite well for problems with little labeled data. You can also find hands-on tutorials guiding you through the details like tutorial 1 or tutorial 2.

For more advanced check out prototypical networks.

Andreas Look
  • 931
  • 5
  • 14