Skip to content

Appendix A. Configuration file description#

The configuration file has a module-based structure. Each liveness algorithm has:

  • basic parameters which are extracted from Liveness::Basic section;
  • a set of unique parameters which are extracted from corresponding Liveness:: section.

Common parameters.

Parameter name Description Default value
maxRuntime Parameter is responsible for the liveness scenario’s length (in frames). If liveness is not verified during amount of frames less or equal to maxRuntime value, liveness scenario is failed. 140
successLength Parameter is responsible for the required length of successful actions in the liveness test (in frames). For example, angle liveness test is passed if amount of frames with angle higher than angleThreshold exceeds the successLength. 5

Parameters for tracking analysis.

Parameter name Description Default value
trackAreaThreshold Sets face detection area alteration threshold. If the difference between detection areas on the previous and the current frame is higher than the frame area multiplied by threshold, track loss is detected. 1
trackCenterThreshold Sets face detection rectangle center alteration threshold. If the difference between rectangle centers on the previous and the current frame is higher than the frame width multiplied by threshold, track loss is detected. 1
allowedTrackLoss Sets the number of frames in a row that can have no track on them. If amount of frames exceeds the threshold, a liveness scenario is failed. 20
verticalPadding Sets vertical padding for allowed area on image. As starting condition face should be placed in formed frame. Result frame height = input image height - 2verticalPaddinginput image height. 0.01
horizontalPadding Sets horizontal padding for allowed area on image. As starting condition face should be placed in formed frame. Result frame width = input image width- 2horizontalPaddinginput image width. 0.01
allowMultipleFaces Allows or forbids multiple faces on image, liveness test will fail, when more than one face will be detected, if this value is set to 0. 1
landmarkThreshold Sets face landmarks alteration threshold. If the difference between mean landmarks alteration on the previous and the current frame is higher than the threshold, track loss is detected. If threshold is set to 0.0 landmark tracking is disabled. 0.0
redetectTolerance Sets the tolerance which is used in redetection. Refer to FaceEngine Handbook (chapter "Face detection facility" section "Face detection") for more information. 20

Angle liveness parameters.

Parameter name Description Default value
angleThreshold Sets the threshold of the corresponding angle which should be surpassed. depends on the liveness type
angleRange Sets the initial range of all angles required to start the test. For example, if roll is +5, pitch is +8 and yaw is -7, and angleRange is 10 the test can be started. 10.0
heldAngleRange Sets the range of not tracked angles, which should be maintained during the test. For example, if test requires pitch angle to be above 20, increasing/decreasing yaw above/less +-heldAngleRange will result test failure . Used in unified liveness. 15.0

Eyebrow, Mouth, and Eye liveness parameters.

Parameter name Description Default value
successThreshold Sets the final threshold of the measured value in corresponding tests. Similar to angleThreshold in angle liveness. depends on the liveness type
startThreshold Sets threshold which is used as starting condition for tracked attribute. In case of eyebrow liveness startThreshold is initialized with value calculated on first frame. depends on the liveness type
fraudThreshold Sets the threshold which is used to track fraud attempts. For example, if test requires pitch angle increase, and calculated mouth distance > mouth fraudThreshold then fraud attempt is detected and test results failure. Used in unified liveness. depends on the liveness type