"Verification" section#

The “Verification” section is intended for creating, deleting, testing verifiers, and editing their parameters. Verifiers are used to quickly compare two faces by face photo image and Face ID, external ID, attribute, event, and display the result of the test (Figure 48).

“Verification” section contains the following elements:

table of verifiers:
“Description”—verifier name;
“Verifier ID”—verifier identifier;
—button for testing the verifier (1);
—button for editing verifier parameters (2);
—button for deleting the verifier (3);
“Add” button—button for creating a verifier;
the number of verifiers displayed on the page is set by the switch in the lower right corner of the page. There can be 10, 25, 50 or 100 verifiers in total on one page (4).

Verifier creation#

To create a verifier, click on the “Add” button (Figure 48). A form for step by step verifier creation will open (Figure 49).

Fill in all the required parameters and click on the “Next” button to proceed to the next step. The description of the parameters is presented below, in the "Verifier editing" section.

After setting all the parameters, a window with a message about the successful verifier creation appears (Figure 50). Click anywhere outside the successful verifier creation message to navigate to the “Verification” section.

Verifier testing#

Testing a verifier is performed by clicking on the button in the line (1 in Figure 48).

The general view of the “Verifier testing” form is shown below (Figure 51).

“Verifier testing” form contains the following blocks:

“Search by”—search options:
“Face”—search by registered face in the system:
“Face ID”—face identifier that is created in the LUNA PLATFORM 5 system as a result of a detection event and attribute extraction;;
“External ID”—search by external face identifier:
- “External ID”—external face identifier;
“Attribute”—search by face attribute:
- “Attribute ID”—attribute (descriptor) identifier;
“Event”—search by registered event in the system:
- “Event ID”—identifier of the event of detection and attribute extraction;
Photo image—search by uploaded photo image:
field for uploading a photo image;
Searching results:
“Photo”—sample of detected face (candidate);
“Similarity, %”—similarity value, in percent;
“Status”—verification result:
- —successful verification;
- —unsuccessful verification;
“Link to the reference” —go to the page the reference face;
—button for downloading the result of the verification (Figure 52).

To test the verifier by face, in the “Search by” block enter the Face ID and select photo image, click on or “Select file”, and specify the path to the image file.

Figure 52. “Verifier testing” form. Search by face

Image file requirements:

*.jpeg, *.png or *.bmp format;
image size no less than 320x250 and no more than 3840x2160 pixels;
image may contain one or more people;
image must have a person's face.

When loading a photo image containing many faces, the Service verifies all faces in the image.

To reset the image, click on .

Verifier editing#

Editing of verifier parameters is performed by clicking the button in the line (2 in the Figure 48). The general view of the verifier editing form is shown below (Figure 53).

Description of the parameters of the verifier editing form is given in the tables (Table 20-25).

Table 20. Parameters of the verifier editing form: general parameters and determined attributes

Parameter	Description	Default value
General
Verifier name \| Specifies the name that will be displayed in the list of verifiers		-
Similarity threshold \| Specifies a similarity score, which will consider that the reference and the \| candidate contain the face of the same person		0.93
Determined attributes
Basic attributes (gender, age)	Assessment of the basic attributes of a person in the image. On When the attribute is enabled, the “Save if” and "Call only in cases" options become available
Head position	Assessment of the head position (angles of inclination and rotation of the head left/right and up/down). When the attribute is enabled, the options “Discard face images with head rotation/tilt angle above” become available	On
Emotion	Determination of the dominant emotion (anger, disgust, fear, happiness, neutral, sadness, surprise)	Off
Mask	Assessment of the presence or absence of a medical mask or mouth covering. When the attribute is enabled, the filter “Process images only if detected” becomes available	Off
Image quality	Determination of quality (the presence of overexposure, blurring, underexposure, the presence of glare on the face, uneven lighting)	On
Eye direction	Assessment of the direction of a person's gaze in the image	Off
Presence of glasses	Assessment of the presence of glasses and their type (glasses, sunglasses, no glasses)	Off
Eye status	Evaluating whether a person's eyes are open or closed in the image, as well as determining key points of the irises of the eyes	Off
Mouth status	Closed or occluded mouth detection and smile detection	Off
Position of 68 feature points of the face	Determination of 68 feature points of the face (requires additional time for calculations, it is used to determine emotions, eye direction or Liveness check)	Off
EXIF metadata	Defining image metadata	Off

Table 21. Parameters of the verifier editing form: Liveness check

Parameter

Description

Default value

Perform Liveness check

Enabling Liveness check

Off

Liveness threshold

Ignoring images with a Liveness score below the specified value.

Possible values: from 0 to 1.

Available only if the “Perform Liveness check” attribute is enabled

0,5

Image quality threshold

Ignoring images with a Liveness quality lower than the specified.

Possible values: from 0 to 1.

Available only if the “Perform Liveness check” attribute is enabled

0,5

Table 22. Parameters of the verifier editing form: Deepfake check

Parameter

Description

Default value

Perform Deepfake check ¹

Determination of digital manipulations for replace one person's likeness convincingly with that of another

Off

Deepfake threshold

Ignoring images with a Deepfake score below the specified value.

Possible values: from 0 to 1, where 1 is a real person, 0 - fake

0.5

Use specified Deepfake mode

Possible values:

Mode 1;
Mode 2;

The choice of mode determines what set of neural networks perform photo processing for deepfake checking.

For more information about the neural networks used in deepfake verification modes, contact VisionLabs technical support.

Mode 2

Table 23. Parameters of the verifier editing form: image quality check

Parameter	Description	Default value
Perform face image quality check	Enabling quality check
Image format	Must be saved in .jpeg or .png format (correct verification). Possible values: JPEG; JPEG2000; PNG;	JPEG; PNG JPEG2000;
Image size in Mb	This assessment determines the size of the image in bytes. It also compares the estimated value with the specified threshold	5120: 2097152
Image width in pixels	This assessment determines the width of the image in pixels. It also compares the estimatedvalues with thresholds (according to ISO or custom thresholds)	180:1920
Image height in pixels	This assessment determines the width of the image in pixels. It also compares the estimated values with thresholds (according to ISO or custom thresholds)	180:1080
Image aspect ratio	This assessment determines the proportional ratio of the image width to height. It also compares the estimated value with the specified threshold	0.74:0,8
Degree of illumination uniformity	It is possible to evaluate the uniformity of illumination according to the requirements specified in the ICAO standard. It also compares the estimated value with the specified threshold (correct verification)	0.3:1
Degree of image specularity	Bright light artifacts and flash reflection from glasses are not allowed (indirect verification)	0.3:1
Degree of image blureness	The pixel colors of front-type photo images must be represented in the 24-bit RGB color space, in which each pixel has 8 bits for each color component: red, green, and blue (indirect verification)	0.61:1
Degree of absence of underexposure in the photo	An underexposure assessment is available. It also compares the estimated value with the specified threshold	0.5:1
Degree of absence of overexposure in the photo	Too much exposure assessment is available. It also compares the estimated value with the specified threshold	0.57:1
Face illumination uniformity	It is possible to evaluate the uniformity of illumination according to requirements specified in ICAO standard. The face should be evenly lit so that there are no shadows or glare on the face image. It also compares the estimated value with the specified threshold (correct verification)	0.5:1
Skin tone dynamic range	This assessment is a determination of the ratio of the brightness of the lightest and darkest areas of the face according to the requirements specified in the ICAO standard. It also compares the estimated value with the specified threshold (correct verification)	0.5:1
Degree of uniformity of the background	This assessment determines the degree of background uniformity from 0 to 1, where: 0 — non-uniform background; 1 — uniform background;	0.5:1
Degree of lightness of the background	This rating determines the degree of background brightness from 0 to 1, where: [0...0.1]—black background; [0.1...0.3]—dark background; [0.3...0.97]—light background; [0.97...1]—white background;	0.5:1
Presence of radial distortion (Fisheye effect)	Possible values: No — the Fisheye effect is not presented in the image; Yes — the Fisheye effect is presented in the image	No
Type of image color based on face	Possible values: Color; Grayscale; Infrared—near-infrared	Color
Shoulders position	This assessment determines the position of the shoulders if they are in the frame: Parallel Non-parallel Hidden	Parallel
Face width in pixels	This assessment determines the width of the face in pixels. It also compares the estimated value with the specified threshold	180:1920
Face height in pixels	This assessment determines the height of the face in pixels. It also compares the estimated value with the specified threshold	180:1080
Face offset from the top edge of the image in pixels	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification)	20:50
Face offset from the bottom edge of the image in pixels	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification)	20:50
Face offset from the left edge of the image in pixels	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification)	20:50
Face offset from the right edge of the image in pixels	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification)	20:50
Head yaw angle	Head rotation should be no more than 5° from the frontal position (correct verification)	-5:5
Head pitch angle	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification). The tilt of the head should be no more than 5° from the frontal position (correct verification)	-5:5
Head roll angle	The image must contain a full front view of the person's head, including the left and right ear (if person has any), the top point of the forehead area and the chin (correct verification). The inclination of the head should be no more than 8° from the frontal position (correct verification)	-8:8
Gaze yaw angle	This assessment determines the direction of gaze (yaw)	-5:5
Gaze pitch angle	This assessment determines the direction of gaze (pitch)	-5:5
Probability of smile presence	The facial expression must be neutral (indirect verification).	0:0.5
Probability of mouth occlusion	It is not allowed to cover the face with hair or foreign objects along the entire width, from the eyebrows to the lower lip (indirect verification)	0:0.5
Probability of open mouth presence	This assessment determines the state of the mouth The mouth is closed (correct verification)	0:0.5
Smile properties	This assessment determines the state of the mouth The facial expression must be neutral (indirect verification). Possible values: None—smile is not found; Smile with closed mouth; Smile with teeth	None
Glasses	Sun glasses are not allowed (correct verification). Possible values: Sunglasses; Eyeglasses; No glasses	No glasses
Left eye status	Both eyes are open normally for the respective subject (considering behavioral factors and/or medical conditions, correct verification). It is not allowed to cover the face with hair or foreign objects along the entire width, from the eyebrows to the lower lip (indirect verification) Possible values: Open; Closed; Occluded	Open
Right eye status	Both eyes are open normally for the respective subject (considering behavioral factors and/or medical conditions, correct verification). It is not allowed to cover the face with hair or foreign objects along the entire width, from the eyebrows to the lower lip (indirect verification). Possible values: Open; Closed; Occluded	Open
Red eyes effect presence	Possible values: No—there is no red-eye effect; Yes—there is a red-eye effect	No
Distance between eye centers in pixels	The image must contain a full front view of the person's head, including the left and right ear (if person has any),the top point of the forehead area and the chin (correct verification) The distance between the centers of the eyes must be at least 120 pixels or at least 45 pixels in accordance with paragraph 12 of the procedure for placing and updating biometric personal data in a unified biometric system (correct verification)	90:100
Horizontal head size relative to image size	This assessment determines the horizontal head size relative to the image size. It also compares the estimated values with thresholds (according to ISO or custom thresholds)	0.5:75
Vertical head size relative to image size	This assessment determines the vertical head size relative to the image size. It also compares the estimated values with thresholds (according to ISO or custom thresholds)	0.6:0.9
The position of the center point of the face horizontally relative to the image	This assessment determines the horizontal position of the center point relative to the image. It also compares the estimated values with thresholds (according to ISO or custom thresholds)	0.45:0.55
The position of the center point of the face vertically relative to the image	This assessment determines the vertical position of the center point relative to the image. It also compares the estimated values with thresholds (according to ISO or custom thresholds)	0.3:0.5
Eyebrows state	The facial expression must be neutral (indirect verification). Possible values: Neutral; Raised; Squnting; Frowning	Neutral
Headwear type	Possible values: None; Baseball_cap; Beanie; Peaked_cap; Shawl; Hat with earflaps; Helmet; Hood; Hat; Other	None
Presence of natural lighting	The face should be evenly lit so that there are no shadows or glare on the face image (correct verification) Possible values: No—the lighting is unnatural; Yes—the lighting is natural	Yes

Table 24. Parameters of the verifier editing form: save parameters

Parameter	Description	Default value
Save parameters
Save image in database	Saving the image in the LUNA PLATFORM 5 database. If enabled, the unconditional saving of images in the database is performed.	Off
Save biometric template in database	Saving the created biometric template in the LUNA PLATFORM 5 database. If enabled, the unconditional saving of biometric templates in the database is performed.	Off

Table 25. Parameters of the verifier editing form: filters

Parameter	Description	Default value
Filters
Discard images with multiple faces	Determination of images containing multiple faces. Possible values: Select only one face of the best quality—process an image containing several faces, but detect only a face of the best quality; Do not discard—detect all faces in the image; Discard—ignore an image containing multiple faces	Do not discard
Reject descriptors with quality below the specified threshold	Ignoring low quality images. To use the filter, you must enable the determination of the descriptor in the determined attributes	0,5
Process images only if detected	Possible values: Missing—the event is created when there is no overlap of the face by the medical mask (no mask); Occluded—the event is created in case of detection of face overlapping; Medical mask—the event is created when a medical mask is detected on the face. Several filter values can be specified. Available only when defining the “Medical mask” attribute	-
Discard face images with head rotation angle (to the left or right, yaw) above	Ignoring images in which the person's head is turned to the left or right at a too large angle —no information will be extracted when detecting a face and evaluating the angle of head rotation. Available only if the “Head position” attribute is enabled	30
Discard face images with head tilt angle (to the left or right, roll) above	Ignoring images in which a person's head is tilted to the left or right at a too large angle—no information will be extracted during face detection and head tilt evaluating. Available only if the “Head position” attribute is enabled	40
Discard face images with head tilt angle (up or down, pitch) above	Ignoring images in which the person’s head is tilted up or down at a too large angle —no information will be extracted during face detection and head tilt evaluating. Available only if the “Head position” attribute is enabled	30
Process images of faces only with Liveness states	Processing images with Liveness status: Spoof—the absence of a “live” person in the frame; Real—the presence of a “live” person in the frame; Unknown. Available only if the “Perform Liveness check” attribute is enabled	-
Process images of faces only with Deepfake states	Processing images with Deepfake status: Fake—the absence of a “live” person in the frame; Real—the presence of a “live” person in the frame. Available only if the Perform Deepfake check” attribute is enabled	-
Filter images based on face image quality assessment results	Filter images according to the parameters set in the "Perform face image quality assessment" setting that comply with ISO/IEC 19794-5:2011 and ICAO Available only when the parameter “Perform face image quality assessment” is enabled	Off

After finishing editing the verifier, click on the “Save” button in the upper right corner.

Verifier deleting#

Deleting a verifier is performed by clicking on the button in the line (3 in the Figure 48).

In the pop-up window (Figure 54), you must confirm the action — click on the “Delete” button or cancel the action by clicking on the “Cancel” button.

Figure 54. Verifier deletion confirmation

Deepfake license required. Deepfake check is not performed on normalized (centered and cropped) images after face detection. ↩