Amazon Rekognition Improves Accuracy of Real-Time Face Recognition and Verification
Amazon Rekognition has improved real-time facial recognition and verification accuracy for even more accurate facial authentication and people search use cases. For face verification applications, Rekognition is now up to 80% more accurate in distinguishing between people who look very similar to each other, and up to 35% more accurate in recognizing the same person with substantial changes in their appearance (hairstyle, hair color, facial hair, glasses, etc.). This reduces both false accepts and false rejects when matching pairs of faces.
With Amazon Rekognition, you can also perform real-time search against tens of millions of faces. With this update, Amazon Rekognition is now up to 25% more accurate in picking out the right face from a digital gallery containing millions of faces.
Amazon Rekognition is helping customers create new experiences for their users. FamilySearch, the largest genealogical organization in the world, uses Amazon Rekognition to help fulfill their mission of connecting families across generations. "FamilySearch developed ‘Compare-a-Face’ using Rekognition to help site users see which of their ancestors they most resemble based on family photographs," said Tom Creighton, CTO and Lead Architect at FamilySearch. "Rekognition was used to provide an engaging experience that helped people relate to their forbearers in a new way."
To try out these updates, simply log in to the Amazon Rekognition Console or create a new collection of faces as described in the documentation. Please note that existing face collections will continue to use the previous face model. For more details on face model versioning, please see this.
This update is available in the US East (Northern Virginia), US West (Oregon), US East (Ohio), EU (Ireland), Asia Pacific (Tokyo), and Asia Pacific (Sydney) regions.