TOKYO–(BUSINESS WIRE)–BTC Corporation (HQ: Tokyo, Japan, CEO: Ken Sugiyama, hereafter BTC) reports that it has achieved the world’s highest level of age error accuracy by AI face recognition using a camera.
In the model that detects a face with a camera and estimates age and gender by AI, the network of face detection and age / gender estimation has been defined separately. BTC has made it possible to incorporate information from the middle layer of CenterNet, which performs face detection, into the age estimation network. This makes it possible to capture information reflected in surrounding people and background images from the information convoluted for face detection, in addition to the features that were previously only for individual face images. We were able to improve the accuracy of estimating the age from the face.
This result is compared with the accuracy of the 2019 paper adopted by CVPR, which is a top-class academic society, and by exceeding the age error accuracy, we have achieved the accuracy that can be called the world’s highest level. For details on the method and accuracy used in BTC, please refer to the preprint posted on arXiv (https://arxiv.org/abs/2010.07544).
About BTC Corporation
BTC is promoting digital transformation of companies with advanced cloud integration services such as cloud, big data, AI utilization, and RPA (Robotic Process Automation). We also have a base in Silicon Valley, USA, and constantly collect the latest information technology and provide it as a solution using the verified technology.
Corporate Planning Office