Face Recognition
The Face Recognition component allows users to identify previously registered faces in real-time images or videos. It compares detected faces with a database of registered faces, using criteria such as similarity, eye distance and eye angle to validate correspondence. It is ideal for applications such as access control, security and attendance recording, among others.
Credits Used
This component takes up 3 credits on the platform, considering a frame rate of 5 FPS and a resolution of 640 x 480 pixels per frame. See section Understanding the Concept of Credits for more information.
Component Parameters
Parameter | Description | Example | Required? |
Component Name | Internal name for the component. Used by our API to select the component for metrics acquisition. | FaceRecognition | Yes. Filled automatically. |
Component ID | Numerical identifier for the component. Used by our API to select the component for metrics acquisition. | 12 | Yes. Filled automatically. |
Name | Descriptive name for the component. | Entrance_Camera | No |
Faces Collections | One or more Faces Collections to which detected faces will be compared to. | Residents | No |
Similarity Threshold | Minimum level of similarity (between 0% and 100%) required to consider that two faces are the same. | 75 | Yes |
Eye Distance Threshold | Maximum distance between eyes, in pixels. | 60 | Yes |
Eye Angle Threshold | Maximum tilt of the eyes, in degrees. | 10 | Yes |
Minimum Face Size | The minimum size (in pixels) of a face to be considered for recognition. | 100 | Yes |
Frequency Threshold | Minimum number of consecutive frames in which the face must be recognized. | 30 | Yes |
Notes
Collections are sets of data (such as faces) that can be stored on the platform and used in cameras. The detected faces are compared with those stored in a collection, enabling the automation of actions such as granting access or triggering security notifications, among other uses. See Working with collections for more information on how to create and manage collections.
The configuration of the thresholds helps avoid false positives by ensuring that recognition is accurate even in varying lighting, distance, and motion conditions.
Eye Distance Threshold is used to validate the alignment of the captured image with an already registered face. If this criterion is not met, the detection will not be taken into account.
Eye Angle Threshold is used to validate the pose of the face. If this criterion is not met, the detection will not be taken into account.
Frequency Threshold ensures that the face is detected correctly by the algorithm. For example, if this parameter is set to the value of 5, the algorithm must make 5 consecutive detections of the same face to trigger a result.
For best results, we recommend a maximum vertical angle of incidence of 15 degrees, and a maximum horizontal angle of incidence of 30 degrees. For more information, see the topic Angles of Incidence in the Camera section in Hardware Requirements for the AIOS Platform.
Metrics and Visualizations
This component does not generate metrics and visualizations that can be displayed in widgets on an application. See sections Working With Applications, Working With Widgets and Metrics and Visualizations of this documentation for more information.