how to implement image Vector search with convex?
What would be the right way to implement using convex stack. I might be wrong but I've check that, openai doesnt have the api for converting image to embeddings. Any suggestions ?
7 Replies
I haven't used Replicate, but they have models for generating embeddings from images, like this one: https://replicate.com/daanelson/imagebind
daanelson/imagebind – Replicate
A model for text, audio, and image embeddings in one space
thanks for the response @erquhart , so i can use this embeddings and store it in convex and use it similar to how i would use text vector search right ?
Yes 👍
hey @erquhart im pretty new with this AI thing, what is
dimensions , the thing i should put in the schema?
how do i know, what size should i put
btw, @erquhart i did manage to make the image search works, its pretty awesome
Nice! If it says it received a vector with 1024, that should be the number to use