Instead of having to do a search on every face, can we look for faces which already found this one as similar? When those matches are stored can they be stored on the foreign face?
Could we also look at strong groupings. IE: all these faces have very high correlations with each other.
All of this comes down to costs:
-Rekog
-Dynamo/storing reverse tables also (and fwd for that matter)
-compute of going through memory, and/or addl complex dynamo queries.
Instead of having to do a search on every face, can we look for faces which already found this one as similar? When those matches are stored can they be stored on the foreign face?
Could we also look at strong groupings. IE: all these faces have very high correlations with each other.
All of this comes down to costs: -Rekog -Dynamo/storing reverse tables also (and fwd for that matter) -compute of going through memory, and/or addl complex dynamo queries.