Posted On: Mar 18, 2019
Amazon Rekognition provides a comprehensive set of face detection, analysis, and recognition features for image and video analysis. Today, we are launching accuracy enhancements to the face analysis features. This is the fifth model update overall since the service launched. Face analysis generates rich metadata about detected faces in the form of gender, age range, emotions, attributes such as ‘Smile’, ‘Eyeglasses’ and ‘Beard’, face pose, face image quality and face landmarks. With this release, we have improved the accuracy of gender identification, emotion detection (for all 7 supported emotions: ‘Happy’, ‘Sad’, ‘Angry’, ‘Surprised’, ‘Disgusted’, ‘Calm’ and ‘Confused’) and attributes such as ‘EyesOpen’. This release is particularly useful for customers who need to search and categorize large photo collections. For example, using face analysis, customers can easily find all photos that contain smiling people, or all photos with men who have beards and are wearing sunglasses.
Woo, a leading dating app in India and for Indian diaspora globally, had this to say: "To rapidly curate new users and ensure that profiles and photos meet our high standards, we were using manual curation to review all profile photos. For a growing business like ours, this approach could never scale fast enough. Amazon Rekognition APIs, particularly DetectFaces, provide rich image metadata. Utilizing information such as number of faces, size of faces, gender and estimated age range from the API, we were able to remove manual curation work entirely. From a lag time of half a day, we are now able to curate images in seconds. Moreover, Amazon Rekognition has kept improving, leading to better overall quality of automatic curation over time". Customers have used our face detection, analysis and recognition features in a variety of use cases such as fighting against human trafficking (Marinus Analytics and Thorn), reducing fraud for mobile payments in underbanked economies (Aella Credit), and streamlining security at concerts (K-STAR Group).
Enhanced face analysis models are now available for both Amazon Rekognition Image and Video, and are the new default for customers in all supported public AWS regions. This feature will be available in AWS GovCloud in the next 8-10 weeks. No machine learning experience is required to get started.