Research on the application of a new generation memory architecture in computer vision AI solutions for IoT devices

Blueshift Memory is seeking to drive the development of a next generation of IoT Computer Vision (CV) application on edge devices. Computer Vision refers to the computer process to analyse visual data as a human would, and make inferences about what that data contains. When integrated in an application, these inferences can be turned into actionable responses.

CV is increasingly being leveraged to solve many diverse real-world problems, ranging from security and health care to manufacturing, smart cities, and robotics.

Bringing a CV application to production requires integrating several hardware and software components. CV deployment relies on cloud communication which is considered a flexible solution for machine learning applications, however mobile data transmission has high latency, is expensive and data in transmission poses a security risk. The alternative and increasingly popular option is to take CV applications to the edge.

Edge or IoT devices are small and lightweight devices which a CV application can be deployed and run.

The market for CV Edge devices is growing fast, in autonomous vehicles, manufacturing, health screening and military / civilian bodycams. However, the current technology with traditional slow processing and high-power consumption prevents the application of Artificial Intelligence (AI) to provide real-time insight from the environment they are interacting with.

Instead they rely on humans or centralised Cloud computing, resulting in latency and poor reliability. In many theatres of operation timely decision making is critical as latency can result in life or death outcomes.

Blueshift Memory's goal is to research the feasibility of a computing device to be used as a primary component for developers to deploy in CV and deliver AI applications on Edge. We expect that a Blueshift Memory design-enabled FPGA component will have a considerably faster calculation speed (5-10 times faster) and significantly lower energy consumption (30-50% reduction).

The potentially lower price of our proposed design would make it possible to leverage existing IoT and CCTV camera infrastructure, providing a vitally important additional functionality.

Blueshift Memory-enabled IoT devices will be able to perform tasks considered impossible in the past, Including weapon or suspicious object detection using bodycams or fixed CCTV. Police officers wearing our enabled AI CV bodycams could be alerted in real-time and the Operation Centre notified automatically when a weapon is detected. Identifying suspicious activity at the edge means only relevant data is transmitted to Central controllers, providing effective direction of resources to hot spots.