Face analysis

id3 Face SDK can analyse a wide range of facial attributes. For instance, it can evaluate the compliancy of a portrait to the ISO/IEC 19794-5 full frontal and the ICAO Portrait Quality Standards thanks to the following modules:

Attribute	Description
Age	Estimated age of the subject.
Background uniformity	Estimated color and structure uniformity of the background in a portrait image. See also Face background removal.
Colorized image	Value indicating whether an image is colorized or grayscale.
Expression	Estimated facial expressions of a subject classified in six universal emotions: angriness, disgust, fear, happiness, sadness, surprise, and additionally neutral.
Eye gaze	Estimated gaze direction for both eyes. When the subject looks directly towards the camera, the return values equal 0.
Eye opening	Degree of opening of both eyes. The resulting scores are in the range 0 to 100.
Face attributes	Group of additional attributes, including: Glasses Hat Make-up Male (gender estimation) Mouth opening Smile
Face mask	Presence of a protective mask on the subject face.
Geometric attributes	Set of geometric characteristics of a portrait image. Resolution: Checks if the image resolution is sufficient. Vertical position: Checks the distance from the bottom edge of the image to the imaginary line passing through the center of the eyes is between 50% - 70% of the total vertical length of the image. Horizontal position: Checks if the head is correctly centered horizontally. Width of head: Checks if the image width is conform to the (image width / head width) ratio of 7:5. The head width is defined as the distance between the left and right ears. Length of head: Checks if the distance between the base of the chin and the crown is less than 80% of the total height of the image.
Glasses attributes	Additional set of attributes related to glasses (if present). It includes: Frame on eye Heavy frame Tinted glasses
Head pose	Estimation of the head pose. Yaw, pitch and roll angles are returned in degrees.
Landmarks	Estimated 68 landmark features
Noise	Quantity of noise in the image.
Occlusions	Presence of occlusions on the subject’s face (eyes, nose, mouth).
Photographic attributes	Photographic quality attributes such as brigthness, dynamic range, flash reflection, natural skin color, exposure, pixelation, sharpness.
Red eyes	Detect if a person’s eyes are red or not (due to flash mostly).

Parameters

Parameter	Description
Expression sensibility	The sensibility of the expression classifier.
Over-exposure sensibility	Sensibility of the over-exposure classifier.
Under-exposure sensibility	Sensibility of the under-exposure classifier.
Thread count	Number of threads to be used for face analysis.

Example

Python

# initialize the FaceAnalyser module
face_analyser = FaceAnalyser(
    thread_count = 4
)

# compute age
age = face_analyser.compute_age(image, face)
print(f"Age: {age}")

# compute landmarks
landmarks = face_analyser.compute_landmarks(image, face)

# compute eye-gaze
eye_gaze = face_analyser.compute_eye_gaze(image, face, landmarks)

# Release all resources allocated to the FaceAnalyser module.
del face_analyser

See also

• FaceAnalyser Class