Supremind (闪马智能) recently closed its Angel Round of funding of CNY 58 million (USD 8.18 million). While some consider it hard for new AI startups to compete in the area of security, Supremind has found a way to survive – and perhaps thrive.
Supremind is a Shanghai-based computer vision specialist founded in 2019 by Peng Yao (彭垚), former founder of the AI Lab of Qiniu Cloud. Based on its self-developed VisionMind big data platform, the startup focuses on video content analysis to help smart city operation and regulation. Currently, it targets five scenarios, including transportation management, large-scale traffic passenger control, urban governance, public service administration and Internet content control.
National security and social stability is China's top priority. China's security surveillance industry takes about 40% of the global market. It has grown from 2007's CNY 145.3 billion to 2017's CNY 648 billion, with CAGR of 16%. The number of Chinese public security companies reached about 30,000 in 2018, with employees numbering around 1.6 million. The output value in 2018 was CNY 660 billion, with a CAGR of 10%. Security products are broadly applied in the ‘Safe Cities’ program, smart traffic, smart buildings, culture and financial areas, accounting for 24%, 18%, 16%, 13% and 12% respectively.
Video surveillance and public security is so far China’s best performing sector in utilizing AI, and it is the first sector that achieves scalability in monetizing AI. Computer vision technology development is the underlying reason behind the mass application. For example, the accuracy rate of image recognition kept improving during the past several years and surpassed human accuracy of 5.1 percent by 2015, as indicated by ImageNet Challenge.
Three types of companies are considered incumbents in this sector: traditional surveillance companies, including Hikvision, Dahua and Uniview who sell edge computing enabled smart cameras, cloud service providers, including Huawei via supporting cloud services, and AI unicorns, such as SenseTime, Megvii, Yitu and CloudWalk, who provides computer vision solutions. Due to the complex procedures to land government deals, it is often believed that new startups have little opportunities in this area.
Supremind is building a competitive edge by leveraging its AI algorithms, lower price and quick deployment. For Hikvision and Dahua, who focus on devices and edge analytics, it allows smart cameras to transmit structured data and simple decisions, such as a vehicle crossing a line, back to be analyzed. It saves storage and bandwidth but the devices are expensive and analysis tends to be simplistic. On the contrary, Supremind could process video streams recorded by traditional cameras from multiple channels. In this way, it could help regulators to make arrangements on a macro level. Besides, the price is much cheaper and deployment time is shorter.
Compared with earlier AI startups, who were dedicated to a smaller number of scenarios with high level of accuracy based on structured data, Supremind targets a broader scope of scenarios using videos. This makes its products more suitable for analysis over a larger geographic region, and a longer time horizon.
Currently, this security newcomer is cooperating with the Shanghai Public Security Bureau. After gaining experience in analyzing the complex cases in Shanghai, it expects to expand to 30 more cities next year. According to Supremind, its estimated revenue in 2019 is CNY 30 million.