(On the left is Huang Zhongsheng, co-founder and CEO of Hao buying clothes, and on the right is co-founder and CTO Chai Jinxiang)
Food, clothing, shelter and transportation almost refer to the basic needs of everyone's life, and "clothing" ranks first. The garment industry is almost one of the most traditional industries in human society, and it is also one of the industries with the least changes from ancient times to the present.
But this does not mean that the clothing industry has not kept pace with the times, good to buy clothes virtual fitting and its co-founder and CTO Professor Chai Jinxiang, want to become the most powerful driving force behind the clothing industry into the AI era.
Just in the new fashion activities that Tmall has just opened, "virtual fitting" has become the biggest highlight of the entire women's clothing venue. Users conducted a deep and personalized virtual fitting interactive experience to achieve immersive virtual fitting with their real face and body.
(In the Tmall new fashion activity, the virtual fitting of good clothes has received a good response)
Clothing retail industry in the Internet era
After entering the era of the Internet and mobile Internet, e-commerce has completely changed the pattern of the retail industry, and has also had a huge impact on the clothing retail industry, and the position of people buying clothes has shifted from the offline part to the online part.
The Internet e-commerce model solves an important problem in traditional clothing sales, that is, to solve information asymmetry. In traditional offline sales, consumers can browse the goods are limited by time and space, but this restriction is lifted online, and people can get more product information.
According to the monitoring data of the China Electronic Commerce Center, the transaction scale of China's garment online shopping market reached 745.7 billion yuan in 2015, and the penetration rate of apparel online shopping was 34.7, while in 2016, it reached 934.3 billion yuan, and the online shopping penetration rate was 36.9%.
In other words, more than one-third of the clothes of The Chinese people are purchased online. Internet + has become a ubiquitous infrastructure like hydropower, directly replacing some links in the business processes of the traditional clothing industry and even creating new business processes.
The dilemma of online clothing retail in the post-Internet era and the attempt at "virtual fitting"
However, this is not a time to sit back and relax.
In the past two years, the Internet traffic dividend has gradually disappeared, for the clothing industry, only to expand the sales channels to Tmall, Jingdong is not enough, and now the major clothing retailers, clothing manufacturers in the enjoyment of the Internet after the traffic dividend, once again facing the problem of slow sales. There are indeed some gaps between online shopping and offline.
The first is that when shopping online, consumers lack effective purchasing decision-making tools. Clothing is non-standard, very personalized products, consumers want to see "the effect of this dress on me" when buying, rather than photos shown by models. Consumers can not confirm the goods, for the merchant, the biggest problem is that the online store compared to the offline store return rate increased a lot.
Second, the online model lacks an excellent user experience compared to offline retail. In retail stores, customers can get the help of shopping guides (although there is a certain promotional nature), matching clothes, and choosing sizes, which is difficult to do in online stores. Although it is possible to consult through customer service, the experience is obviously greatly reduced.
As Luo Zhenyu said, in the post-Internet era, in the case of the disappearance of traffic dividends, how to seize traffic time and obtain traffic conversion capabilities has become the biggest challenge for everyone.
In fact, the apparel retail industry is also trying accordingly. In the past, the industry has been talking about the new methods provided by new technologies such as "three-dimensional fitting" and "VR/AR fitting". There are some new things such as two-dimensional texture matching, three-dimensional virtual fitting room, virtual fitting mirror and so on.
Although virtual fitting startups around the world are working hard, including Fits.me, Metail, UNIQLO, etc., are the pioneers of virtual fitting, they are currently facing problems such as technological stagnation and business expansion difficulties.
(Second generation virtual fitting technology: 3D virtual fitting room renderings)
To some extent, these technical means are difficult to solve the following problems. The first is the large-scale application, conventional technology is difficult to ensure the efficiency of virtual fitting, and the cost of its use. The second is the sense of integration of virtual and real, if the consumption in the 3D model, to meet the real, beautiful, personalized fitting needs, is a big problem. Finally, the Internet and the purchase path, the purpose of virtual fitting, fundamentally speaking, is to sell clothes, then its application must produce a reasonable purchase closed loop, and mobile use mode.
Good Buy clothes helps fill the gap in online fitting through AI technology
However, like many other industries such as finance, automobiles, and home furnishing, some problems have been well solved after the emergence of artificial intelligence. Professor Chai Jinxiang founded the virtual fitting of good clothes to buy clothes, is to solve the problems existing in the virtual fitting and even the entire clothing manufacturing and retail industry through machine learning.
Professor Chai Jinxiang said: "Many companies' virtual fitting products only focus on the fitting effect, but the body reconstruction is the most basic part. "A virtual fitting scene that can land on the ground can not only restore the details, materials, drape, etc. of the clothes, but also produce a consumer who is close to the real consumer himself."
Using basic information such as height and weight entered by users, checking body characteristics, etc., combined with its exclusive female body data database, thousands of body figure data can be deduced. Machine learning and graphics can be used to model the human body more accurately for consumers. Finally, the facial photo taken by it is reproduced in three dimensions, and it can get a consumer "self" that is close to the real one.
(Good to buy clothes intelligent reconstruction of three-dimensional mannequin, error value of only 1-1.5 cm)
(Good to buy clothes intelligent three-dimensional face reconstruction effect)
The effect of virtual fitting is mainly manifested in two aspects for online clothing retail.
On the one hand, the user's dwell time increases. In the general online store, the user stays for about 2 minutes is a relatively normal time, but after testing, after using the good buying virtual fitting, the user will generally stay for more than 10 minutes or even longer.
On the other hand, it is the user who enhances the purchase confidence through virtual fitting. With virtual fittings, customers can more easily understand what clothes they are more suitable for, so their purchases are more targeted, and with body data, the size of their purchases is more appropriate, so it is not easy to frequently return and exchange goods.
In the just-started Tmall Fashion Festival, Good Buy analyzed the warm-up period of the past 3 days: it was found that the per capita visit time was as long as 23 minutes, and more users were online for more than 50 minutes; 34.7% of the total users clicked on the Tmall baby details page of the item during the fitting process, and 15% of users directly and decisively clicked to buy, compared with the daily baby detail page of the original flagship store, the purchase rate was only 5%.
In addition to "fitting", Haoyi hopes to provide more artificial intelligence support for the clothing industry
Professor Chai Jinxiang said that the virtual fitting is actually just a beginning, while the virtual fitting, good clothes will get more suitable for consumer data, including body data, preference data, size matching data, etc., using these data, there is hope to have a fundamental impact on the clothing industry. In particular, there are some parts of the industry that are in urgent need, such as pre-sale, shopping guide, personalized customization, smart plate play, clothing design, etc.
Clothing pre-sale
The pre-sale model is a new business model produced by the clothing retail industry after entering the Internet industry, and it is also a new attempt, which directly determines the production volume by the consumer's willingness to buy, solves the problem of inventory backlog in the traditional clothing industry, and directly links the production end with consumers, thereby improving the user shopping experience, not only the clothing brand omits a lot of intermediate links, pre-sale is also a unique "advance consumption" experience for users in the future of online shopping.
However, similar to online shopping for clothes, the pre-sale model also has the problem of whether the size matches the customer and whether the style is appropriate, so after using the virtual fitting, the brand can further solve the problem of the size ratio in the pre-sale and the problem of user return and replacement in the future.
Smart shopping guide
In high-end offline stores, and some high-consumption groups, shopping guides not only assume the role of guiding customers, but also provide customers with fashion consulting services and clothing suggestions. Professor Chai Jinxiang revealed that a senior Stylist (fashion consulting) consulting has a repurchase rate of more than 80%, and will greatly improve the conversion rate of customers.
However, its high-end attributes are destined to be difficult to achieve at scale. Chai Jinxiang said: "It must be emphasized again that clothes are non-standard, worn on others to look good, worn on their own is not necessarily good-looking, to recommend suitable clothes to users, especially need to consider the user's appearance, body, temperament, style, offline shopping guide can do, but for brand online sales, is a big lack." ”
Good buying artificial intelligence technology can solve this problem, using exclusive user body and face data, accurate clothing labels, other collected data, and machine learning to analyze consumer information and preferences, and in the future can also provide customers with intelligent shopping guide services. Professor Chai Jinxiang also revealed that the smart shopping guide for good clothes currently has an early Demo, which is likely to be launched for consumers to experience in the first half of this year.
Personalization
Personalized customization, that is, the C2M model, can tailor clothes for consumers to suit their body characteristics, appearance, and preferences, which is an area that many clothing brands want to try. But at present, it can only be achieved in categories such as men's suits and shirts and in a very small number of high-end women's clothing.
The reason is very simple, in suits and shirts, the style of clothing is single, only the color, size, etc. need to be changed, and it is easier to design and produce in the case of maintaining clothing models. However, the category of women's clothing is complex, and it is difficult to complete the process of patterning and production in a short period of time. And whether it is men's and women's wear, this C2M model is basically unable to achieve scale.
With the help of artificial intelligence, personalization is possible, the customer's body information is measured, the body model is reconstructed, and then combined with its physiognomy characteristics, the design and typing process is optimized. In addition, the whole process is basically completed by the algorithm, and the labor cost is reduced, and the large-scale application can be realized.
The smart man's desk matches the size
Professor Chai Jinxiang said that the original specification of the four sizes of S, M, L and XL currently used in China came from Japan in the 1980s. Putting aside the gap between Chinese and Japanese consumers, in the past 30 years alone, the figure of domestic consumers has undergone tremendous changes, so this standard size and the human table used in the clothing production process actually need to be adjusted accordingly.
A large amount of data on the customer's body shape obtained through virtual fitting can provide intelligent desk services for clothing manufacturers in the future, which can not only meet the reality of current consumers, but also adjust the size details of the table according to the different objects of clothing.
Costume design
In fact, in the future, in addition to customized clothing, clothing manufacturers will also have their own regular launch of new clothing, so a large number of user data feedback can also help them to design and manufacture clothing.
In the current garment industry, there are still information barriers between manufacturers and consumers, although they will conduct market research and statistics on the sales of previous products, but this part of the information is only macro data such as shipments and sales. As for who the clothes are sold to, and whether they are the same as the target customers at the time of design, the clothing manufacturers do not have an answer.
In the future, good clothes can provide more comprehensive customer preference information for the clothing design link through big data, so that manufacturers can aim more accurately in the design and truly hit the "key point" of consumers.
As mentioned above, the goal of Chai Jinxiang and Hao buying clothes virtual fitting is not only to stay on the "fitting", which will be an opportunity for a comprehensive change in the clothing industry.
Similarly, artificial intelligence has already brought us changes, but it will bring us more surprises in the future. As Chai Jinxiang said: "The development of technology is actually not a step-by-step development, but at a certain stage there will be large steps and breakthroughs, and it is still difficult to accurately predict the future." ”
It takes more of our imagination.
The author of this article, Li Jixiang, columnist of Yiou; WeChat: ligi_avd (please indicate "name - company - position" for convenient remarks when adding); please indicate the author's name and "source: Yiou"; the content of the article is the author's personal opinion, and does not mean that Yiou agrees or supports the views.
reward
65
86