During the development of the application: https://openexchange.intersystems.com/package/companies-search and our studies on vector search, we did not explore image vectorization, which involves transforming unstructured data beyond text, including images.
We are unsure if this is possible due to a lack of knowledge. However, if we can achieve image vectorization, we could revolutionize image processing by enabling the search for word similarities within images. For example, if we want to find companies with covered parking and they have photos of the parking lot, image vectorization could enable the API to return these companies.
Thank you for posting the idea. According to information from experts, this functionality already exists.
I don't see that anything specifically would be needed in IRIS
This look like a some work with embeddings, depends on the exact task, look for embeddings model, or prepare own model, IRIS can store Vectors, and do operations with it, that's already enough