Amazon detects fear with biometric facial analysis as MoodMe partners with Texas A&M

The biometric face analysis capabilities of Amazon Rekognition have been upgraded with improved accuracy and functionality, including detection of fear, according to a company announcement.

The inclusion of fear brings Rekognition’s detection capability to eight different emotions, and accuracy of detecting any of them has been improved, the company says.

Face analysis also generates a range of metadata, including detection of gender, age range, attributes such as “smile,” face pose, face image quality, and face landmarks, in addition to emotion. The age range estimation feature has been upgraded to provide narrower age ranges for people in most age groups.

Amazon also updated Rekognition to enhance its facial analysis accuracy earlier this year.

A team of researchers led by Texas A&M University Department of Computer Science and Engineering Assistant Professor Dr. Zhangyang “Atlas” Wang is collaborating with brand engagement and facial analysis company MoodMe to improve its facial analysis and recognition algorithms for video conference participant re-identification and emotion and attention detection, Texas A&M Today reports.

Wang and his team of students research machine learning and computer vision, and will develop algorithms to identify participants in video conferences across different sessions and dates, regardless of different clothes, hairstyles, lighting, and seating arrangement. By integrating deep learning into the system design, they plan to improve facial analysis, but also the accuracy of emotion detection.

Previous research projects by Wang and his team have caught international attention and been published in top publications, according to the report, and Wang says he is happy to tackle a real-world challenge for an in-demand application, and to work with a leading company like MoodMe.

MoodMe builds embedded software for facial insights and augmented reality face filters. The company has worked with some household name clients, like At&T and FIFA, and according to Texas A&M Today, respects the privacy of consumers while helping customers engage them and analyze their experiences.

“Privacy is at the center of all our work, both in research and in product engineering,” comments Chandra de Keyser, CEO of MoodMe. “All the insights we gather from faces fully respect people’s privacy – no faces are stored ever, nor sent to the Cloud. With all the promises of internet giants, there is still no “one-click” to delete all our pictures or data that they have accumulated about us. The focus of our research and engineering is to reach the highest performances and precision on edge computing platforms like smartphones, embedded/internet of things, robots and desktop computers.”

A team of academic researchers recently found that biometrics-based emotion detection is not reliably accurate, with people making expected facial expressions only 20 to 30 percent of the time.

Article Topics

accuracy  |  Amazon  |  analysis  |  biometrics  |  emotion recognition  |  facial recognition