What kind of future retail?
When Amazon Go, the first unmanned convenience store in Amazon, was officially open to the public, it was discovered that by using technologies such as sensor fusion, visual identity, and artificial intelligence, shopping without cash, without queuing up and holding the desired product straight out could be realized Experience.
Retail merchants embrace new technologies to provide consumers with time-saving and new experience has become the consensus of Chinese and foreign industries.Capital think no one retail, simply emphasize 'no one' is not the key value is around the needs of users, improve the C-side Shopping efficiency, meet the needs of subdivision scenarios, changing consumption patterns, while helping to improve efficiency and return on B side.
And this kind of experience without paying for a queue is also the direction of the efforts of Chinese technology companies.
AI enhances retail experience
While Amazon Go's main highlight is 'No Checkout,' the retail industry is not interested in mass marketing 'no one.' "The growth in unmanned stores in China is questionable, not so much market space and hard to find New store, 'said Chen Haibo, founder and CEO of Deep Blue Technology.
Founded in Shanghai by the Australian Returning Team, Shenlan Technology has more than 60 PhDs, postdoctoral fellows and masters from prestigious universities both at home and abroad. It has made substantial achievements in convolutional neural networks, deep learning, machine vision and biometrics. It has won In the declaration of the patents, software copyrights more than 50, the products are sold to the United States, Australia, Japan and New Zealand and other 17 countries and regions.
In February 2017, Silan jointly launched a takego 'take-away' settlement-free payment system in conjunction with Ant Financial Ltd. In June, Unmanned Retail Kiosk, AI Self-Contained Shelves, AI Self- Traders market, in fact, is one of the scenes without retail.
However, Chen Haibo clearly realized that 'consumers do not need no retail, businesses need unattended retail. C-side regardless of the B-side pain, C-side does not care (care) B-side of the United States.' As a traditional retail smart upgrade solutions Providers, he admitted that many unmanned retail solutions currently do not really improve consumer shopping efficiency, but he is optimistic about the use of technology to liberate the clerk devoted more to communicate with customers.Therefore, he hopes to quickly cut into the market, Do stock market upgrade.
To this end, deep blue technology to computer vision and biometric identification and payment programs applied to the people, goods, field identification.
In the field of computer vision, we use body recognition algorithm to track and locate people with the camera, while the commodity area is composed of several cameras working together to set up the electronic fence line, cross-validation algorithm, real-time to the background algorithm, real-time contour algorithm to complete the goods Precise identification This process requires real-time adjustment of the weight of each camera to process data in order to prevent the effect of 'Blind Touch', while in the settlement area will determine whether the customer took the goods away and eventually generate bills.
This algorithm and calculating power have very high requirements.On the commodity area to determine the location of goods, there are several background and the relationship between the goods need to be considered.Fixed background and fixed commodity identification is the lowest technical difficulty.In the fixed background, In judging mobile products, multi-level multi-frame collaborative error correction algorithm needs to be adopted to cross-identify and judge the goods through multiple cameras, and the hardest part is to judge the infinite background and infinite goods.
Chen Haibo told CBN reporter that at present, three hours of machine training are required to complete the comprehensive recognition of the commodities. Data and models are crucial in machine training. "The key to this method lies in GPU computing power. Big data Of the processing power.Using the camera to shoot at 120 frames per second, behind the multi-frame multi-machine error correction algorithm. "Chen Haibo said.
In addition, there are 3D body detection biometric detection technology for user identification and biological payments. Deep blue technology will be based on artificial intelligence technology to facilitate the transformation of the convenience store program, the United States to build self-service enterprises thousands of open-door machines , To achieve the sweep hand (face / code) opened the door and took it away to promote the development of self-service aircraft in the Chinese market. January 2018, deep blue technology began to use artificial intelligence technology into the office near-field retail, has introduced the market for office shelves Near-field retail scene upgrade products.
For tech companies that want to make a difference in the retail industry, there are three main points in 2018: first, an upgrade of office shelves; second, an upgrade in the auto-expo sector; and third, retail and mobile.
Commercialization is the key
The future of online and offline integration of new consumption patterns is the trend of the times.Considering the hustle and bustle of investment ventures, technology and commercialization is the key.
Nomura China Telecom and technology industry analysts should be heavy Hee in the interview with the first Financial said that the current image recognition applications in the retail industry is still not large body mass, while facing multiple challenges, the algorithm is relatively high, such as energy No identification of twins, articles, such as the use of openings and then put back will affect the actual promotion.
'Speech Recognition, Image Recognition and other technologies are used in many industries, but not very fast large-scale cash opportunities,' he said.
Hua Ying Capital Investment Director Jiang Zhifeng accept CBN reporter interview also stressed the importance of commercialization.He is optimistic about the deep blue technology as ToB enterprises, with the real strength of the technology will be able to algorithm and technology into Adapt to different scenarios, different needs of products, and mass production.
Today, Deep Blue Technology has strategic cooperation with a number of brands, retailers and e-commerce providers including Ant Financial, Wahaha, Meiling and Micro Union.
'Deep blue technology as an artificial intelligence unattended retail solution provider, has a set of AI recognition algorithm specifically for retail scenarios, which is the core value of deep blue technology.' Jiang Zhifeng said.
Further from the B end, unattended retail as a new enhanced channel for the brand to provide a new line of traffic; the other hand, the original traditional offline channels can not accurately identify consumers, smart retail terminals you can understand consumer precision portraits , Based on a set of membership system, to achieve consumer closed-loop upgrade.
Referring to the outlet of the retail industry, Jiang Zhifeng believes that the new retail is large and complex, there are many segments, whether it is the optimization of large-scale supermarkets or consumer research, shelf monitoring has operational space. Is a small convenience store, whether it is Ali, or other start-up companies, but also some people are concerned about how to enhance their efficiency.With the constant maturity of technology and consumer habits, intelligence, facilitation, no Humanization is the inevitable development of the industry.
On the 5th, artificial intelligence company deep blue science and technology announced the completion of 100 million yuan level A round of financing, capital led by Hua Ying Capital, Germany capital with the vote.This is the second harvest of Angel Wheel and Yunfeng Fund tens of millions of Pre-A After financing, deep blue technology for the third time in a year access to capital favor.
CPT Capital also hope that the development perspective of the unmanned retail industry, not simply confined to the current unmanned shelves, unmanned containers and other forms, products and models will continue to iterate, can be applied to a variety of scenarios. 'No one Shelf has its advantages, but also has the challenge, in the item of theft, control point of view, we think it is not yet reached a mature stage.We believe that the uncertainty of unmanned shelves, need a suitable environment, it Certainly a future direction of technology, but you can not face the challenges ahead, which is one of our investors focus on the core issue. 'Jiang Zhifeng said.