Hello friends!
I am trying to replace TinEye in my project. ( a photo matching system with an already existing reference image database)
In general now SIFT gives more or less results. But I need to optimize the search in the database. Otherwise comparison with 10 thousand pictures will take forever. Now I look towards FAIS and start to study vector compression algorithms.
Perhaps you can tell me how normal people solve this problem?
“content based image/media retrieval”
browse the scientific literature. you’ll find it’s a lot.
it involves various ways of clustering.
it might involve data structures for optimized retrieval, or it might just involve linear scan over reduced-dimension vectors (or signatures).
it might involve range queries, or KNN queries, or any other type of query.
these days, you might use AI to generate the feature vectors, rather than dumb old local feature descriptors.
the “build” tag does not apply to this topic. I have removed it.
possibly related (posted within hours of each other):