Image similarity

Image similarity is a way to quantify how similar 2 images are.

Image similarity is a way to consistently quantify the similarity between images. Image similarity search is fast since the data becomes so compressed.

Similarity is a vague concept when humans get to decide. Deep learning image similarity makes it easier to consistently quantify the similarity between images.

Leather jacket similarity
Figure 1. Are these images similar? Sure, why not.

What can you do with image similarity?

If your criteria for “similarity” are visual but not well defined, or if you lack finely labeled data for training, you should use image similarity with deep learning to give you a boost.

Reverse searching

Reverse searching lets you start with the content and gives you the associated content. Here are some examples:

  • Display alternative products in case of items are out-of-stock or sold out

  • Locate duplicates entries on digital assets and bank of images for removal or deduplication

  • Identify the best-matching products, models, or designs from a repository of referenced images

Similar chairs

Data clustering

Image similarity can be applied to data clustering. This allows you to leverage a combination of explicit information, like the type of clothes, and visual features that are learned implicitly by a deep learning model.

Clustering retail stuff
Figure 2. Clustering items from an online retail shop. Categories like the shoes are visually unique, while some categories have a fair amount of visual overlap with their neighbors.

Similarity implemented on the platform

This is how image similarity is implemented on the platform:

  1. First, you use a deep learning model to convert each image in your existing dataset into an embedding vector. You take the vectors from an Output block.
    All images’ vectors are saved into a large table, the Index.

  2. To find similar images to a new image, you transform that image into a vector with the same deep learning model, then you compare that vector with every vector in the index to pick the most similar ones.

How image similarity search works

Tutorial - Find similar images of fruits

If you want to try out image similarity on the platform, you should do our tutorial Find similar images of fruits. In this tutorial, you’ll learn how to build, train, deploy and use an image similarity model.

Image similarity cheat sheet

We’ve created a cheat sheet for how you work with similarity on the Peltarion Platform. This cheat sheet will guide you quickly through the main steps you take when solving your problem.

Deep dive explanation of image similarity

If you’re interested in deep diving into the technical side and get an idea of how image similarity actually works, this is an excellent post by Peltarion employee Romain:

What is similar to a zebra?

Was this page helpful?