Our team has built a platform that allows shops, restaurants, and other public venues to gather information on their visitors using cameras, connected to the cloud. The system utilizes powerful facial recognition algorithms to detect whether a person has been seen before, their age, and other demographic information. Combined with point-of-sale integration, the system can provide priceless real-time insights for employees, such as a person's past purchases, or recommendations on their purchases. Business owners can obtain detailed reports on customers, shopping profiles, and various metrics. The system can operate even on consumer-grade IP cameras and does not require any servers, which makes integration easy and inexpensive. The system is optimized for performance, and can complete the full cycle of identification and recommendation faster than it takes for a person to enter the shop.
Behind the scenes, the system runs on the AWS cloud and consists of a video processing fleet, a machine learning cluster, business logic servers, and a content delivery infrastructure. The algorithms are based on cascades of machine learning models and streaming analytics. All processed data is encrypted both in transit and at rest, and depending on local laws, the system can operate on impersonal data. Its streaming nature and reliance on the cloud allows easy scaling up and down, which drastically reduces the operational costs due to the nature of retail, which mostly active during daytime.
