In 2014, I traveled to Xichang for the first time. Qionghai is unforgettable in its picturesque scenery, but what impresses me even more is the people who live in Daliangshan. At a market, I chatted with a Yi sister who sold souvenirs. She told me cheerfully and proudly that she had walked fifteen hours up the mountain to get into town. I still remember the bamboo basket she was carrying, and underneath it was a two-year-old child, and she was wearing straw shoes on her feet.
I remembered a place name that day, and she told me she was from Meigu.
In the work of poverty alleviation, there is a so-called "China's poverty alleviation looks at Liangshan, and Liangshan's poverty alleviation looks at meigu". A few years passed, and in November 2020, Meigu County successfully lifted itself out of poverty. In the context of comprehensive poverty alleviation, Meigu County has now solved the problem of eating and wearing. But I still think of Daliangshan from time to time, and I think that the eldest sister's child should have already gone to elementary school. What kind of education is he receiving? Can he share the future of technology's rapid development like children in big cities?
Recently, I had an opportunity to see the children in Meigu County through the lens, and also learned about their encounters with AI and cutting-edge technology. Not long ago, HUAWEI CLOUD came to Xiaoming Aixin Primary School in Caihong Village, Meigu County, Daliangshan, and brought a vivid "AI lesson" to the children here. Through the lens, we can see the smiling faces of the children together and the moment when they are together with AI technology.
What we want to talk about today is the gift that HUAWEI CLOUD gave to the children: a warm, wear-resistant, and beautiful winter coat.
This dress not only carries the heart of HUAWEI CLOUD and the warmth of the children in Liangshan, but also sews in the new clothes, about the development of ai models, and about the deterministic future of industrial intelligence.
New clothes and hope in Liangshan
At the end of 2021, Ren Xing, a senior AI researcher at HUAWEI CLOUD, received a call from an old friend of his. This friend's name is Jordan, he was once a senior student at Peking University, and now he is the principal of Xiaoming Aixin Primary School in Caihong Village, Meigu County, Liangshan. Principal Jordan hopes that HUAWEI CLOUD can bring an AI lesson to the children. Ren Xing, who gladly accepted this invitation, is also ready to bring more gifts to the children with the relevant team of HUAWEI CLOUD.
After much deliberation, the team decided to give the children a brand new "AI winter coat". Although the living environment of Daliangshan is no longer what it used to be, the winter here is very cold, the children are growing up, and they are in the age of sports, and it is always good to have more practical clothes. Therefore, HUAWEI CLOUD hopes to send children a winter coat that takes into account the attributes of warmth, wear resistance, dirt resistance, and beauty, and can also wear more than one winter coat with more practical value. At the same time, the team also wanted to integrate some Yi ethnic elements into this dress to express good wishes for the children.
However, in the traditional clothing design model, it is very challenging to meet so many needs in a short period of time, because the cycle of customizing a large number of ready-to-wear garments is generally very long. HUAWEI CLOUD quickly came up with a way to get AI and human designers to complete the design of clothes. This can not only complete the "extreme operation", but also perfectly fit the theme of letting AI enter Daliangshan this time.
As a result, HUAWEI CLOUD contacted Shidi Intelligence, a cutting-edge company dedicated to empowering the fashion industry through technology. The designers of Shidi Intelligent joined hands with the HUAWEI CLOUD Pangu model to break the traditional design cycle of three weeks. The children's winter coat design was completed in just one week, thus guaranteeing this special gift for the children before the New Year.
A new piece of clothing may not be expensive, but the expectations it carries at this time are so heavy. The growth of these children is the hope of Liangshan. When they come into contact with AI, when they understand technology, maybe we can hear the silent gear turning, that is the hope of intelligent China.
Designer's assistant and partner
It may not only be the children of Daliangshan, but more people will be curious about how the designer collaborated with the AI to complete the new clothes design.
Here we first need to extend a little industry background: many people think that fashion and clothing design should be an industry that pays attention to inspiration and craftsmanship, but is some distance away from cutting-edge technologies such as AI.
But in fact, empowering the fashion industry and the apparel design industry with digital and intelligent technology has become a new trend and industrial blue ocean. For example, Shizhen Intelligence, which completed the design of this winter coat for the children of Daliangshan, is such a company. They want to improve and enhance the way the fashion industry is designed, manufactured, produced and presented through the latest technology. In terms of technology selection, Shidi Intelligent has built the industry's first virtual 3D prototype digital platform for footwear, bags and building, providing digital solutions for the fashion industry with AI, big data, intelligent three-dimensional design CAD, extremely realistic rendering and other technologies.
In the field of AI, Shidi Intelligent has diversified cooperation with HUAWEI CLOUD. For example, on the basis of the Pangu model, a dataset for clothing categories is established to help designers, for example, combined with HUAWEI CLOUD's AI capabilities, industry AI applications such as rapid preview of clothing design are constructed.
How exactly did AI capabilities help designers complete the design of Winter Clothes for Children in Liangshan - in response to this question, Lin Zisen, CEO of Shidi Intelligence, gave the answer.
Contrary to popular imagination, fashion designers are not an industry that relies entirely on inspiration and imagination to get work done. In real fashion design work, designers have to go through a very long planning process. First of all, we must target the audience of the crowd, and secondly, we must clarify the design needs. And these workflows have a lot of tedious, repetitive, and complex parts, which have the potential to be handed over to AI to complete.
For example, in the process of designing winter clothes for children in Daliangshan, designers want to design patterns that conform to the meaning and are beautiful and generous, they need to find a large number of materials, and then think about inspiration on the basis of materials. This process was originally long and cumbersome, but with the help of the Pangu large model 'Wensheng Tu' ability, designers can enter text to generate the big model across modes and directly give the pattern that meets the needs. On the basis of the design reference of the large model, the designers save a lot of time in finding materials, and can focus more on the core design links.
After determining the pattern of the fit, designers need to make physical samples to consider whether the final combination is appropriate. With the support of HUAWEI CLOUD's computing power and data, Shidi's intelligent AI clothing design system can change the material, color, and clothing pattern of the layout map online, which can be completed in just a few minutes. Through the real-time rendering technology of Shidi Intelligent, 3D samples can be selected online, avoiding the process of repeated proofing. With the help of these AI capabilities, the designers finally completed the design requirements for sending winter clothes to children on the basis of two or three hundred design schemes generated by the Pangu multi-modal large model.
From the design process of new clothes, we can see that cross-modal, high-efficiency AI capabilities have abundant value in the apparel industry, especially in the design process. Traditional clothing proofing needs to go through a series of steps such as planning, designing style drawings, determining the flat version, preparing ingredients, making samples, modifying samples, taking pictures, setting samples, verifying prices, and issuing process lists, usually taking 17-30 days. With the joint efforts of HUAWEI CLOUD and Shidi Intelligence, the design time can be shortened to 5-7 days.
Behind this winter coat is not just the innovation of AI and human designers. It also shows the deterministic future of AI becoming a designer assistant and partner, and the deterministic future of the value of the industry empowered by the big model.
Insights and value of big models
As can be seen from the case of designing winter coats, multimodal cross-cutting tasks such as "wensheng diagram" appear more and more frequently in various industry fields. In order to complete these intelligent tasks with high efficiency and high level, multi-modal pre-training models have become a window for the industry and AI. Being able to comprehensively use a variety of modal information to understand, think and reason about things, so as to achieve more advanced intelligent applications, is not only the core ability of the multimodal big model, but also the common expectation of thousands of industries for general intelligence.
The HUAWEI CLOUD Pangu multimodal model can accurately realize this value. On the basis of the Pangu large model that can complete AI tasks across modes, only need to be fine-tuned in the fashion industry data, you can quickly get the "Pangu fashion multimodal model" suitable for the apparel industry.
Behind such a playful name is such an industrial value trend: if the industry wants to build a pre-training model that can act on production and improve industrial efficiency, it needs to cross a huge computing power threshold and accumulate heavy research and development capabilities. The reality is that a large number of industries that are far away from AI cannot afford these costs, and intensive training of cross-modal large models, and then industrial landing in the form of fine-tuning of industry data sets, is the most reasonable way to land large models at present, and it is also the core channel for the industry to rapidly improve its intelligent capabilities.
Let's go back to the field of fashion design and see what value the Pangu model can bring in this industry space.
First of all, the Pangu big model can help designers gain insight into fashion trends, analyze elements such as colors, layouts, patterns, etc., and then visualize fashion elements to designers, so as to improve the designer's insight ability and insight efficiency.
Secondly, the Pangu model can diversify the generation of costumes, support the use of Wensheng diagrams, and generate suitable materials and ideas for designers according to their basic needs. This greatly improves the efficiency of designers in collecting materials and accumulating inspiration, so that designers can enter the secondary creation stage faster.
In addition, large models can also help designers generate clothing pictures that meet the requirements in batches, so as to avoid the problems that designers often face such as cumbersome information and inaccurate search results. Avoid designers from repeatedly searching and comparing data, so that the creation company is more accurate and fast.
On the basis of fashion industry data, Pangu big model can carry out a variety of auxiliary design work, ultimately shortening the industry process and improving enterprise efficiency. At present, China's fashion and apparel industry has reached an industrial space of 80 billion yuan. During this period, a large number of design work can be empowered and assisted by AI, which can reduce comprehensive costs and improve industrial efficiency with the help of large models.
Let's once again see the big picture in the small, the fashion design industry is far from design to assist an AI demand. AI fitting, AI generation of model pictures, intelligent communication, AI industrial scheduling and other capabilities have a wide range of industrial needs, AI has a lot of use in the "fashion circle". This is the industrial space that HUAWEI CLOUD AI capabilities can penetrate, and it is the direction in which the Pangu model shows its skills.
And in China and even the world, how many similar sub-sectors are there? Based on the model of big model + industry data fine-tuning, it can be adapted to various industries efficiently and industrially, and quickly bring intelligent value to the specific industrial space. This is the future of AI and the core change of industrial intelligence.
On this day, the children of Daliangshan wear winter clothes that AI designers and human designers have jointly completed.
On this day, designers feel the productivity revolution brought about by the big model and industrial intelligence.
On this day, the children go out of the mountains to explore and build a place called Smart China.
This winter coat guards the hopes of the children of Daliangshan, and perhaps also pins on the hopes of intelligent China.