Since 2016, the sweeping wave of "unmanned retail" has caused people in the artificial intelligence industry to stop and watch.
"How accurate is the identification? How to avoid the risk of theft? How to track and identify technically? How to choose products" and other questions have all faced AI technology companies that focus on retail.
As the topic of unmanned retail gradually heats up, SandStar, an artificial intelligence company that empowers the retail industry with AI technology, was born in October 2016. SandStar firmly believes in the great empowerment of AI to the traditional retail industry, and is committed to using artificial intelligence to connect technology and commerce, and using the deep integration of computer vision and AI big data to promote the digital and intelligent upgrade of retail. At present, SandStar has matured 4 production lines, namely smart container, smart store analysis, pure visual unmanned store and AI settlement desk.
In 2020, the sudden new crown epidemic will have a huge impact on China's manufacturing, catering and retail industries. In order to control the continuation of the epidemic, the State Council took the lead in extending the Spring Festival public holiday. At the same time, the government is also advocating to avoid gatherings and reduce contact between people. "Non-contact consumption" appears to be particularly important in the isolation economy, and virtually no one is born. Huge demand for sales and new retail.
Take the smart container for non-contact consumption as an example. At present, the number of domestic vending machines is about 500,000. With the changes in social structure, China’s automated vending equipment will increase in geometric progression in the future. Way to grow. The market capacity is expected to exceed 1 million units, and the market scale will reach more than 40 billion yuan.
According to reports, compared with Japan and the United States where smart vending machines are developed, my country's smart vending machine market penetration, utilization, and functional services are far from insufficient. On average, every 25 people in Japan have a smart vending machine, and the average annual consumption per person is about 2,200 yuan. On average, every 48 people in the United States have a smart vending machine, and the per capita annual consumption is about 900 yuan. And nearly 7,277 talents in China will have a smart vending machine, and the per capita annual consumption is only 5 yuan, which is far below the level of the United States and Japan. It can be seen that my country's smart vending machine market penetration still has a large room for improvement.
But for the huge unmanned vending market, is the focus really on "no one"? According to SandStar, the essence of the unmanned vending market lies in how to achieve "cost reduction and efficiency enhancement" through technical means. "Unmanned" is only the natural external performance after the intelligent upgrade, so the Vision Smart Vending Cabinet has always been adhering to the convenient unmanned sales concept of "take it and go, automatic settlement".
When the unmanned sales were on the rise, the core identification technology was also in a state of "a hundred flowers contending", such as RFID, gravity sensing, static identification, etc., and SandStar uses computer dynamic vision technology as the core identification solution. In fact, how to choose the core technology solution, the nature of unmanned retail, the answer will naturally come out.
Take the gravity-sensing container as an example. Due to the working principle of the sensor, all products must be placed in accordance with the original display diagram, and the action of consumers selecting products after opening the door is unpredictable. Faced with a wide variety of product categories, consumers are prone to Repeatedly take and put back the action, once the product is placed incorrectly, it will seriously affect the identification, and the weight of many products is very close, resulting in insufficient data collection granularity, and other identification technologies are required for normal application;
static data The identification requires a camera to be installed on each floor of the container and leave a sufficient height, otherwise the camera view is blocked; the RFID tag solution not only has to bear the high one-time tag cost and labor cost, but the tag is easily torn by people, resulting in a cargo damage rate of over 10% . Therefore, from the perspective of data drive and cost reduction and efficiency enhancement, computer vision dynamic recognition is a better choice.
Shida’s computer dynamic vision container recognizes not only goods but also consumer taking behavior, so the recognition accuracy is greatly improved. In practical applications, firstly, there is no need to formulate a display matrix for product display according to product height and weight, so the replenishment operation is simple and convenient; secondly, the location of consumers' picking and placing products does not affect identification, and the space in the cabinet can be used 100%, which improves operational efficiency.
The other partyOn the other hand, SandStar's dynamic vision solution can also detect surrounding traffic flow, collect consumer behavior data for analysis, and help formulate more appropriate product selection and pricing strategies, which can help increase sales by at least 20%. At present, the computer dynamic visual container of Star Vision is widely used in gas stations, convenience stores, shopping malls, office buildings and other areas. At the same time, SandStar has not only become the common choice of the world’s three largest beverage suppliers, but also cooperates with Unilever and supermarkets. Brand owners conduct in-depth cooperation.
The application of unmanned retail technology is actually very extensive, and it is not limited to smart containers. It has great potential for unmanned stores and data collection empowerment. During the installation of the unmanned store project, SandStar is based on computer vision dynamic recognition and high-level action semantic recognition technology to accurately identify consumer identity behavior trajectories, product handling and other shopping behaviors, and automatically generate bills after consumers leave. Settlement. Even if multiple people enter the store and are blocked, the shopping details can be accurately identified at multiple points to avoid theft and damage. Big data analysis based on deep machine learning realizes smart operation and smart decision-making, improves management efficiency for retailers at a lower cost, and provides consumers with a more convenient shopping experience.
SandStar is to promote computer vision dynamic recognition technology to the market, and actively participates in industry forums and exhibitions. The NRF US Retail Show (one of the world's three major retail exhibitions) that has participated in also brought the first batch of customers to SandStar. . In the past two years, SandStar has also been conducting content operations through self-media and website media, and has established an industry leader position through user cases.
At present, through the polishing of the long-term business model, SandStar can provide traditional container customers with after-installed computer vision dynamic recognition cameras and servers and other hardware product services, and efficiently modify traditional containers; at the same time, SandStar also cooperates with container manufacturers Carry out in-depth cooperation, carry out the pre-installation matching of computer vision dynamic recognition technology, and provide complete smart containers for customers in need. Therefore, the corresponding hardware product fees and software service fees are collected as a profit method. According to reports, SandStar achieved tens of millions of RMB in revenue in 2019 and is expected to reach a balance of payments this year.
Since the establishment of the company, the team size has reached more than 200 people, of which technical research and development personnel account for 90%. According to reports, most of the company’s core team is of Tsinghua background. Wu Yili, founder and CEO of SandStar, was the first batch of graduates from the School of Software of Tsinghua University. An important foundation for the integration of scientific research and business. After graduation, he joined Oracle and IBM in charge of enterprise software business, and accumulated rich experience in company strategy and operation, business management, etc.
At the financing level, SandStar recently announced the completion of hundreds of millions of RMB in Series B+ financing. The investor in this round is the True Digital Group, a digital branch of True Corporation, a leading company in the Thai telecommunications industry and a subsidiary of Thailand's Chia Tai Group. SandStar has completed nearly 100 million yuan in Series B financing led by Guopeng Capital and co-invested by Mobai Capital in September 2019; at the end of 2017, it announced the completion of tens of millions of yuan in Pre-A round financing, and Kunzhong Capital led the investment. Old shareholders Baidu Ventures and Fengshang Capital joined in the investment; in October 2017, it was announced that Baidu Ventures and Fengshang Capital had tens of millions of yuan in angel round investment.