Hailo, an AI startup, has recently announced its collaboration with Foxconn & Socionext to launch an edge computing processing solution, BOXiedge, for video analytics.
The solution could deliver market-leading energy efficiency for artificial intelligence inference, which benefits applications such as industrial IoT (internet of things), smart medical, and smart cities. It is the successor of a mini server launched in January, through the collaboration of Foxconn with Network Optix, which shares the same name. Unlike its predecessor, the new solution can perform pose estimation, image classification, and detection, among other tasks, on footage from over 20 cameras simultaneously. This can be attributed to SynQuacer SC2AA chip of SocioNext and Hailo-8 processor of Hailo, which consumes less power as compared to chips from rivals while incorporating software control, a heat-dissipating design, and memory.
Resources including compute blocks, control, and memory are distributed throughout the whole chip in the Hailo-8 processor. The software of Hailo analyzes each AI algorithm requirement as well as allocates the appropriate modules. Hailo-8 is capable of 26 TOPs (tera-operations per second), which works to 2.8 tera-operations per second per watt. It processed 672 frames per second, with 224 x 224 pixels per inch as the image resolution and sucked down 1.67 watts.
Startups AlStorm, Flex Logix, Xnor, Graphcore, Quadric, and Esperanto Technologies are developing various chips that are customized for the AI workloads. For example, Alibaba, a Chinese retail giant, has launched an artificial intelligence inference chip for logistics verticals, smart cities, and autonomous driving in the 2nd half of 2019.
Foxconn is one of the first, publicly disclosed customers of Hailo after NEC & ABB Technology. The startup has recently stated that it has been working on developing Hailo-8 into products from tier-1 automotive companies and OEMs in various fields such as ADAS (advanced driver-assistant systems) and industries such as smart homes, smart cities, and robotics.
Source credit: