I am working on a face recognition problem where objective is to to reduce the search space for a query embedding to match with the registered embedding. I have 100000 face embedding in total, checking the similarity for a query embedding from the whole 100000 embedding is a time consuming task. So, I used DBSCAN clustering to cluster the 100000 embedding into n clusters, so that I can reduce the search space (for eq: if 100000 emedding are split to 10 clusters of 10000 I need to check only one cluster and thereby I can reduce the search space from 100000 to 10000. The problem I am facing now when a query embedding come for face recognition how to find a probable cluster from the n clusters where the registered embedding of the same person resides.
The problem I am facing now when a query embedding come for face recognition how to find a probable cluster from the n clusters where the registered embedding of the same person resides.