Text-to-Image and Image-to-image search Using CLIP

Search, No Filter
1

Industries today deal with ever increasing amounts of data. Especially in retail, fashion, and other industries where the image representation of products plays an important role.

In such a situation, we can often describe one product in many ways, making it challenging to perform accurate and least time-consuming searches.

This is a companion discussion topic for the original entry at https://www.pinecone.io/learn/clip-image-search/
2

In this tutorial, how would it be possible to replace the pandas dataframe index by a Faiss index ? Even a Flat one. I’ve tried it locally but I’m always getting terrible results when trying to query the text embeding on the faiss index, while the method with the dataframe shown in the tutorial works amazingly.

Thanks for the ideas !