Jiwei Network News, on December 17, at the first Dishui Lake China RISC-V Industry Forum, Huang Qunhui, COO of Beijing Crystal Vision Intelligent Technology Co., Ltd., shared the RISC-V-based AI vision chip "CR182x", which will be released in March next year.
Huang Qunhui pointed out that the goal of Crystal Vision Intelligence is to build the most secure data and the most open artificial intelligence chip platform in the ecological chain. He said that artificial intelligence is divided into three stages, the first stage is server + algorithm, the second stage is GPU + algorithm, both of which face the problems of high cost, high power consumption and difficult deployment. The third stage of the dedicated chip + built-in algorithm can promote the implementation of large-scale data structure.
Jingshi intelligent layout of the security monitoring field is one of the current mainstream applications of AI, the company just released a consumer product at the beginning of this year, next year's upcoming RELEASE OF CR182x series chips will further help it achieve a comprehensive product layout.
According to Huang Qunhui, the CR182x series chips use the dual-core RISC-V IP of flat-headed brother, as well as the TPU and ISP developed by Crystal Vision Intelligence, providing a compact design of dual camera + screen display + built-in DDR in a small package size of 9X9 square millimeters.
The CR182x series chips adopt the flat-headed brother RISC-V C906 core, which is a light and intelligent AI vision chip positioned in the pan-security market. The chip has 0.5TOPs computing power, supports 2-way MIPI or DVP video input, the maximum resolution is 2880 * 1620@30fps, built-in DDR and 100 megabit MAC PHY, is one of the most integrated products in the industry.
The CR182x series integrates the second generation of fully self-developed deep learning tensor processor (TPU), fully self-developed intelligent image processing engine (Smart ISP), hardware-level highly secure data protection architecture (Security), speech processing engine and H.264/265 intelligent codec technology.
The CR182x series adopts fully self-developed professional security level image processing (ISP) technology, which can support 5M HD dual sensor inputs. Each input supports image processing functions such as starlight-level noise reduction, wide dynamics, de-fogging, distortion correction, etc., providing high-quality image output for camera applications. The H.264/265 codec can also be combined with ISP/TPU technology to enhance or retain the key information in the video screen, providing a high-definition and high-resolution low-stream intelligent coding solution.
About Crystal Vision Intelligence
Founded in 2019, Crystal Vision Intelligent Technology Co., Ltd. is a chip design company focusing on video surveillance and edge computing technology research and development, with the domestic performance of leading AI TPU computing core and SoC chip integration technology, the team elite from top universities at home and abroad and top chip design companies, the overall human quality of more than 80% with doctoral or master's degree, team members experience from product design, research and development, production to assist customers with product landing have rich practical experience, The average practical experience of the chip design team is more than 16 years, and it has complete experience in chip research and development, platform architecture design, algorithm development, software and hardware reference design, etc.
At present, Jingshi Intelligent has obtained investment from Xiaomi Yangtze River Industry Capital and Bitmain and other institutions. According to Huang Qunhui, the company has artificial intelligence TPU and image processor ISP full-stack independent research and development technology, and its first CV183x series of professional-grade security chips have exceeded one million sets in the cumulative sales of smart visual surveillance and strong AI scene application markets.