4 ms·
Have you looked at CLIP? You can use that to create a vector embedding for an image that includes semantic information (what's actually in the image - animals,
by simonw 4y ago
Have you looked at CLIP? You can use that to create a vector embedding for an image that includes semantic information (what's actually in the image - animals, colours, etc) - those could then be used with pgvector to find similar images.
CLIP is how search engines like https://lexica.art/?q=6dc768e2-7a7c-494d-9a39-fd8f27e69248 https://lexica.art/?q=6dc768e2-7a7c-494d-9a39-fd8f27e69248 work.