Shanghai Municipal Health Agency to Use Facial/Object Recognition Tech to Ensure Restaurants Comply with Food Safety laws

Remark Holdings, Inc., a global digital media technology company, announced its KanKan subsidiary has been awarded a seven-figure contract by one of the largest state-owned enterprises in China to provide a facial and object recognition technology to a Shanghai municipal health agency, which will use the product to ensure that restaurants comply with local food safety laws.

The technology, which will first be installed in 200 restaurants before later being expanded to 2,000 facilities, utilizes artificial intelligence to ensure that food service workers are wearing hats and masks (as required by the Shanghai municipal health agency) by instantly analyzing images obtained from an artificial intelligence device designed by KanKan and attached to cameras installed in a restaurant kitchen. After analyzing the images, the software detects any violation of the specified food safety laws and extracts screen images, with violation details, that the health agency can then review.

Remark Holdings' uses "deep learning", a type of algorithm-based machine learning that is used to model high-level abstractions in data, to train its artificial intelligence products. Utilizing KanKan's extensive data sets to train its KanKan Artificial Intelligence Platform with tens of millions of supervised and unsupervised samples allows Remark Holdings to develop models that extract facial features and recognize objects, such as branded logos, animals or license plates, with a high degree of precision. KanKan's facial-recognition and object-recognition technology, for example, has demonstrated an accuracy rate of more than 96 percent.

The KanKan Artificial Intelligence Platform is designed as a one-stop shop that will provide small to large enterprises and developers with the ability to customize and train their own artificial-intelligence models for their businesses, with stable and ready-to-use artificial-intelligence modeling stacks and pre-trained models. The platform has been successfully used to train models in high-precision facial recognition with more than 99.63% accuracy (ranked top 15 in the world), and in negative content filtering for topics such as pornography, terrorism, politically-sensitive information and more.