“Pre-warehouse” of daily premium fresh cloth stores realizes extreme speed distribution, Jingdong Home moves physical retail stores to online one after another, netease strictly creates cost-effective goods through ODM mode… As one of the important scenarios for the landing of artificial intelligence, new retail reconstructs the form of retail industry and attracts more and more attention from artificial intelligence enterprises.

In 2016, Ma Yun Qi conference, the first proposed the concept of “new retail”, said that in the future “e-commerce” will eventually be replaced by “new retail”.

With digitalization as the core feature, relying on big data and artificial intelligence technology, the new retail mode has changed from “goods – shop – people” mode to the people-oriented “people – goods – shop” mode. Ma Yun believes that the new retail represents the online and offline all-channel through, combined with modern logistics to form a complete system of the new industry form.

Alibaba’s Hema has become a pioneer in new retail. Hema stores can not only serve as sales stores that attract offline consumers and guide traffic to online, but also serve as distribution and storage bases, which can quickly identify online sales orders, accurately sort goods, transfer goods from the background to the front desk in the first time, and then deliver goods from the stores to users.



Alibaba’s Hema has become a pioneer in new retail

In order to give consumers a better user experience and facilitate the management of various products, big data, cross-camera tracking, live detection and other artificial intelligence technologies run through every link of the new retail.

On the basis of fully understanding the new retail scene, Datatown designs and develops 3D face, live detection, human image matting, human clothing segmentation, cross-camera tracking and other data to support the needs of the scene such as identity authentication, customer emotion analysis, fashion recommendation, customer shopping trajectory analysis and so on.

Identity check

In the new retail stores, the identification of new and old customers and members, self-service settlement and “face brush” interaction are the application of face recognition technology.

Taking Hema as an example, relying on Alibaba’s rich consumer data, Hema obtains clear and comprehensive consumer portraits through member registration, store face recognition, associated user behavior records and other methods.



“Face brush” payment in Hema Fresh Store

In order to ensure the accuracy of face recognition and identity authentication and improve customer experience, many AI enterprises have invested a lot of money to build face databases. For example, Baidu has established a face database for offline store customers, and determined the identity of members and repeat customers through face recognition, so as to carry out accurate customer group management.

Dataston has developed a high standard of face recognition and live detection data set for the identity verification segment of the new retail. When the customer purchases, the data can help to get through the face, ID number, bank card information, online automatic verification of identity information, to achieve the authoritative ID card real-name authentication.

Customer sentiment analysis

By using facial recognition technology to analyze customers’ emotions, new retailers can get a sense of how much they like different products.

Samsung had previously filed a patent with the World Intellectual Property Office containing a deep learning-based algorithm to identify the user’s facial expressions when looking at the device. The patent will be used in Samsung’s offline retail stores.

The algorithm can recognize at least six different expressions, including surprise, happiness, happiness, anger and sadness, and finally make recommendations based on the emotions reflected in the user’s facial expressions.



AI empowers Samsung’s new retail stores

For emotion analysis, Datatang has developed high-standard expression recognition data in the industry. Based on the training of big data, the machine can accurately capture and identify the changes of customers’ facial expressions when they choose products, and realize personalized recommendation.

Fashion is recommended

Now, the recommendation system has become an important traffic entry, who can do better than the user understand the user, who can occupy the initiative of the new retail era.

Alibaba’s Fashion AI project is a new retail attempt to use artificial intelligence as Fashion matching recommendation. For consumers who want to try more creative styles, the system provides cross-brand clothing matching suggestions.

The model behind the project is mainly based on Alibaba Ecosystem’s insight into consumer shopping trends and Taobao stylists’ data of more than 500,000 sets of matching schemes.



The Fashion AI concept store demonstrates a new model of digital Fashion retailing

In the process of automatic fitting, the machine can capture the data of the head picture, figure and other parts, and automatically add the selected clothes to the virtual customers, saving the time of fitting, and the wearing effect is clear at a distance.

Behind the fashion recommendation is the scene technology based on the key points of human body and clothing segmentation. The body key point data and clothing segmentation data developed by Data Hall can help fashion recommend more widely applied.

Data hall owns copyright of human body key points and clothing segmentation data

User shopping trajectory analysis

The machine can further understand the shopping habits of consumers through the analysis of customers’ moving track and length of stay, and can also guide the shopping mall to determine the rent of the store intelligently based on the shopping habits of most customers.

Tencent Mall Smart Retail System is a smart store product based on the support of computer vision and big data, which is committed to realizing the digital transformation of offline stores.

Tencent YouMall has landed in Belle International’s Shenzhen store. After a statistical analysis of the re-ID (pedestrian re-identification) and area heat from an AI camera installed in the store, the system found that customers were more interested in and focused on the women’s leisure area than the other areas.



The application scene of Tencent excellent Mall

Datapot’s copyrighted re-ID and cross-camera tracking data help generate the length of a customer’s stay at a particular location, the movement of the store, and the thermal map of customer flow. According to the results of such statistical analysis, brand owners can optimize the product display and shopping guide reception in stores.

As a new scene of artificial intelligence application, new retail can be improved and perfected in many aspects. This puts forward higher requirements for machine training data sets, and makes the research and development of rich and high-quality training data integration a common direction for AI enterprises.