NIST sets new accuracy threshold for entry into 1:N face recognition evaluation

Starting November 30, 2026, developers seeking to enter the 1:N evaluation must submit at least one algorithm to the 1:1 track.

The 1:N evaluation measures identification, in which a face is searched against a gallery of identities.

Its announcement does not say existing 1:N participants must qualify retroactively.

New datasets for 1:N NIST is also adding multiple datasets to the 1:N evaluation.

Unlike the accuracy prerequisite, which applies to new 1:N participants, the updated speed requirement applies to submissions in both evaluation tracks.