This is the base positives dataset for the NearID project. Each sample contains multiple views of the same identity rendered in different backgrounds/contexts.
Near-identity distractors (different but similar instances in matched context) are available as separate datasets listed below. Together, they form the NearID training and evaluation benchmark.
