GoldWaterFall / Chat-Health

chat health - empowering health management through clinical AI
Apache License 2.0
3 stars 1 forks source link

Appropriate evaluation criteria for face matching #11

Open YuSun7543503 opened 10 months ago

YuSun7543503 commented 10 months ago

In demo3, I add the face image on the id card to enhance the accuarcy of face matching, however, since the photo on card mght be taken before several years, the matching result might be influenced, I'm wondering how much level that we can trust on the matching result based on card photo?

YuSun7543503 commented 10 months ago

The code for matching is:

compare_faces( known_face_encodings, face_encoding_to_check, tolerance=0.4)

Tolerance is the criteria to represent tht difference between 2 faces, I choose 0.4 in the project, may we need more experients to measure the best value for practical use.