|Better Face Detection in Amazon Rekognition|
|Written by Lucy Black|
|Friday, 23 November 2018|
Updates to Amazon's face detection software have led to a significant increase in its performance in detecting faces from images, avoiding false detections and in making matches.
When we put Amazon and artificial intelligence together in the same sentence we are usually referring to speech recognition and Alexa in particular. However Amazon's AI Services also includes Amazon Rekognition which recognize objects in photographs.
Intended to make it easy for devs to add image analysis to apps, it is a fully managed service for image detection and recognition that uses deep neural network models trained by Amazon. Part of Amazon Web Services (AWS), Rekognition integrates directly with Amazon S3 and AWS Lambda for building scalable, affordable, and reliable image analysis applications.
Last time we looked at Rekognition it had added age estimation to its facial recognition abilities. The latest update brings improvements in its ability to detect faces within images along with the ability to perform higher accuracy face matches and obtain improved age, gender, and emotion attributes for faces.
As Ranju Das and Venkatesh Bagaria explain on the AWS Machine Learning Blog:
In real-world images, various aspects can have an impact on a system’s ability to detect faces with high accuracy. These aspects might include pose variations caused by head movement and/or camera movements, occlusion due to foreground or background objects (such as faces covered by hats, hair, or hands of another person in the foreground), illumination variations (such as low contrast and shadows), bright lighting that leads to washed out faces, low quality and resolution that leads to noisy and blurry faces, and distortion from cameras and lenses themselves. These issues manifest as missed detections (a face not detected) or false detections (an image region detected as a face even when there is no face).
They also provide examples of these challenging situations showing how the updated service succeeds in detecting faces which would previously have been missed. The pictures on the right have bounding boxes whereas those on the left don't.
Blur and occlusion:
The bottom line is that Amazon Rekognition can now detect 40 percent more faces and the rate of false detections is reduced by 50 percent. In addition, face recognition now returns 30 percent more correct ‘best’ matches (the most similar face) compared to the previous model when searching against a large collection of faces.
or email your comment to: firstname.lastname@example.org
|Last Updated ( Friday, 23 November 2018 )|