Testing

Testing Face Matching and Face Recognition

  1. Preparing experimental dataset

    1. Determine positive and negative test cases

    2. Positive test cases consist of pairing data, that is, face photos that are matched have the same face photos.

    3. Negative test cases consist of non-pairing data, that is, photos that do not have matching face photos in positive test cases.

  2. Prepare face photo conditions according to real conditions, for example:

    1. Normal face photo A face photo that is the same as point a using certain attributes, such as wearing a cap, wearing a hijab, wearing glasses.

    2. A face photo that is the same as point a with additional changes to the face, such as having a beard, mustache, makeup, thickened eyebrows.

    3. If possible, extreme face photo data can be added, such as: face in different age ranges, face with physical changes such as thin and fat.

  3. Calculation of Metrics

    1. Calculate FAR (false acceptance rate) | FAR = FP / (FP + TN)

    2. Calculate FRR (false rejection rate) | FRR = FN / (FN + TP)

    3. Calculate EER (Equal Error Rate)

PreviousMetricsNextPeople Analytics

Last updated