Neuromorphic vision systems developer Prophesee and partners Digital Media Professionals (DMP) and Restar Electronics Corp. (REC) are teaming to develop an edge AI FPGA machine vision platform based on Prophesee’s Metavision Event-Based Vision technology. The collaboration will also include integration and data collection services. Traditional frame-based machine learning techniques require huge data sets and computing power, and they have accuracy limitations due to the challenges of various lighting conditions and backgrounds. Prophesee develops event-based vision solutions that address issues caused by lighting conditions and background through sparse, continuous event input, enabling enhanced performance in how machine learning can be applied to computer vision. This system will include an event camera using an event-based vision sensor, like Sony’s IMX636ES, which was realized in a separate collaboration. Prophesee recently launched an ultralight, compact HD evaluation kit for the Sony IMX636ES HD stacked event-based vision sensor. Appearance of the DMP ZIA C3 and Prophesee's event-based sensor camera. Courtesy of Prophesee. Event-based vision sensors feature pixels that are each powered by their own embedded intelligent processing, allowing the pixel to activate independently when it detects a relative change in illumination intensity. As event-based vision relies on independent receptors to collect essential information only, it functions much like the human eye and brain working together to overcome inherent limitation of conventional machine vision. The main target use case for the jointly developed FPGA platform is in smart city applications for traffic monitoring, foot traffic, customer flow inside buildings, subway station safety, and smart building. Prophesee has launched a third collaboration, with iCatch, a provider of security surveillance solutions, on event-based metavision sensing projects that integrated iCatch’s V57 AI vision processor with the stacked event-based vision sensor IMX636.