Interpret Video Feed Overlays
Video feed overlays are available in the Camera Feed Analyzer window. Depending on the View menu options you choose, various supplemental information is displayed in real time in the form of overlays and readouts to help monitor the video feed.
The following colors are used to indicate the level of recognition for a face:
- Gray: Unrecognizable. Either the face does not meet minimum quality values to be recognized or the response from the attempted recognition has not yet arrived.
- Purple: Stranger. The face met sufficient quality for recognition but was not recognized and did not meet minimum quality to be registered.
- Cyan: Identified as a close match to recognized user but not 100% identification.
- Blue: Registered person without a name. The face was recognized as matching one already in the Person Directory.
- Green: Registered person with a name. The face was recognized as matching one already in the Person Directory.
- Yellow: Concern. The registered face has been tagged as a concern.
- Red: Threat. The registered face has been tagged as a threat.
The indicator icons indicate the following:
- Gender
- Age
- Sentiment
- Sentiment score
- Smile
The following information is provided in the Desktop Client video display:
- FPS: The number of frames per second being captured by the camera.
- Video: Video resolution. (e.g. 1280x720)
- Detection: Face detection resolution.
- Detector: The detector CPU capacity level. It's based on the number of CPU processing cores.
- DPS: Frame detections per second.
- dDt: Detection time. For example, how long it is taking to detect a face.
- This time should normally be in 20-50ms range.
- dRt: Recognition time. For example, how long it is taking to recognize a face.
- This time should normally be in the 60 - 250 ms range for only identity recognition.
- If age, gender, and sentiment are also being recognized, the time can be expected to be up to 450 ms.
If your processing times are longer than indicated here, it may be an indication your system is overloaded. You should look to offload some of your video feed processing to other computers.
There are 3 main image quality metrics:
- Q: Center pose. Represents how directly the face is looking at the camera. A face looking directly at the camera would receive a score of 1. The more the face looks up, down, left, or right of the camera, the more this metric is reduced.
- S: Face sharpness. Represents how clear the image is. A score of 1 represents a perfectly clear image. while 0 represents an extremely blurry image.
- C: Face contrast. Represents the color contrast within the image. A score of 1 represents an image with very high contrast, while 0 represents very low contrast.
For guidance on these metrics, see Image Quality Metrics Guidance. These metrics can be configured in the recognition preferences menu.
- Face size: Face size is the resolution of the image with a 25% margin. It can be used to ignore background (smaller) faces or to require an up-close high resolution image to be presented before the face is learned by the SAFR system. See detection preferences and recognition preferences for information on how to customize SAFR behavior based on the detected face size.
- Gender: Displayed as an icon if gender recognition is enabled.
- Age: Age of user if age recognition is enabled.
- Face recognition image submission: The thumbnail in the lower right corner is the image submitted for facial recognition. This is only displayed if you select Recognition Candidates (for macOS) or Detection List and Recognition Details (for Windows) from the View menu.