CASTKOM FACE AI KEY HGHLIGHTS
Check the likelihood that two faces belong to same person
Detect human faces in an image, including small, sidewaysand blurry faces. The service can also make machinelearning-based prediction of features and attributes,including age, emotion, gender, pose, smile and facial hairalong with 27 landmarks for each face in the image.
The face API now integrates emotion recognition, returningthe confidence across a set of emotions for each face inthe image such as anger, contempt, disgust, fear,happiness, neutral, sadness and surprise
SIMILAR FACE SIMILAR
Easily find similar-looking faces. Given a collection of facesand a new face as a query, this API will return acollection of Similar faces.
Organize many undefined faces together into groups,based on their visual similarity
FACE AI BENEFITS
- Enables security professionals to shortlist, target and identify intruders, loiterers and potential miscreants
- It also identifies attributes, which include face landmarks such as noses and eyes, gender, age, and other machine-predicted facial features including age, emotion, gender, nose, smile and facial hair, along with 27 landmarks for each face in the image
- Emotion recognition such as anger, contempt, disgust, fear, happiness, neutral, sadness and surprise
- Automatically creates a log of people in the camera or scene
- Allows people to be tracked across multiple cameras or locations to identify potential issues or threats
- Enables automated matching against a watch list, with real time alerting
- Increases productivity and efficiency of security professionals
- Face recognition accuracy up to 95% on public standard data set
FACE AI FEATURES
- Detect, recognize and log faces. Face resolution 70 pixels min,100-300 recommended. Min 30 pixels between eyes, 50-150 recommended.
- Angle variations from frontal face – yaw +/- 25 degrees, pitch +/- 25 degrees, roll +/- 20 degrees
- Crisp, no-blur images required
- Face database allows up to 100K faces
- Multiple enrollment pictures required per person. Video enrollment can be used which captures 5-10 pictures per enrollment.
- On the spot enrollment also supported Recognition in less than 1-2 seconds
- Watch list to save and match in real time
- Alerts provided on watch list matches
- Input: RTSP/HTTP stream from network/IP cameras
- Supports integration with VMS such as Milestone, Pelco
- Works with color, B/W and IR cameras
- Report provided on faces recognized
- Can provide processed RTMP stream with overlay for faces and names if face is recognized
- Search by name or similar faces