Image Similarity Search using CLIP

Build a image similarity search using CLIP and Upstash Vector

This tutorial will show you how to build an image similarity search functionality using Upstash Vector and CLIP.

Clip is a zero-shot image classification network. It’s used classifying an image given possible class names. For example it can choose between dog, cat and plane when shown an image.

But it can also be used to generate embeddings for these images. Since the model will generate similar embeddings for similar images, we can use this to build a similarity search.

Here the outline of the tutorial:

1- Install dependencies and create an index

2- Load the model and dataset

3- Generate embeddings for the images

4- Query the index

5- Outro

You can find the full tutorial and code in the notebook here.