Microsoft Cognitive Services Face API - Identify vs. FindSimilar

Question

I am building a celebrity matching facial recognition service. I have multiple pictures for each celebrity, and I want to match the user's query face to the closest celebrity. There are two ways to do this using Microsoft's Cognitive Services Face API:

• FaceList - FindSimilar to find the closest single face
• PersonGroup - Identify to find the closest person (making use of all faces)

Both can be used with 0 threshold (include all results), but which one would be more accurate? I am concerned whether multiple pictures in different light conditions, with different hairstyle, etc. used together will help or trip the detection accuracy. Thanks!

Answer 1

It depends on what you want to do, if you want to return a collection of similar faces to the user based on their query, then go for Face - Find Similar, but if you just want to identify the people in a photo and return the possible identified person then Face - Identify would be the right choice.

Using multiple pictures in different light conditions, different hairstyles, and etc. for training will only increase the accuracy.