Maintained with ☕️ by
IcePanel logo

Amazon Rekognition launches eye gaze direction detection in Face APIs

Share

Services

Today, AWS announced the general availability of eye gaze direction detection in Rekognition’s Face APIs to support accessibility and safety, validate photos, and help identify where users focus on the screen. The new EyeDirection attribute in Amazon Rekognition DetectFaces and IndexFaces APIs predicts a person’s eye gaze direction yaw (rotation on vertical axis) and pitch (rotation on horizontal axis) angles for each face detected in an image. Using eye gaze direction, customers can improve their applications’ usability by understanding if users focus on desired sections of the application user interface. They can also improve their applications’ accessibility for users with physical mobility limitations through eye gaze enabled navigation. Customers in automotive and fleet management can improve driver safety by detecting if drivers are looking away from the road. Advertising customers can analyze ad creatives to identify the optimal presentation of a model’s eye gaze in highly effective ads. Further, companies and government entities can validate the quality and compliance of their users or employees profile photos. The EyeDirection attribute predicts a value between -180 to 180 degree for the yaw and pitch angles along with a confidence score between 0 and 100\. EyeDirection returns a high confidence score when eye gaze direction can be confidently predicted (i.e. a person’s eye is clearly visible), and a low score when eye gaze direction cannot be confidently predicted (i.e. a person’s eye is closed or not visible). Eye gaze direction detection is now available in all [AWS regions](https://docs.aws.amazon.com/general/latest/gr/rekognition.html) supported by Amazon Rekognition Faces. To get started, download the latest [AWS SDK](https://docs.aws.amazon.com/rekognition/latest/dg/sdk-general-information-section.html) and visit Amazon Rekognition Faces [documentation](https://docs.aws.amazon.com/rekognition/latest/dg/faces.html).