Alibaba Enters The Textile Industry: Converging 200 Thousand Scientists To Create " Textile Brain " Leading The Textile And Garment Industry To Change.

Before digitalization, fabrics are the means of production of sunshine group. After digitalization, fabric and data are the source of real value creation.

The fabric industry itself is a pattern based industry, which makes artificial intelligence based image recognition technology promising in this field.

How to search for needles from more than 80 thousand varieties of fabric materials?

How to find tiny defects in the fabric of ten thousand units?

From the perspective of resources, data is a new means of production for the sun, and its ultimate value is to maximize the value of other means of production.

This is a view expressed by Dr. Wang Jian, the chairman of the technical committee of Alibaba group and the honorary mayor of Xuelang Town, when he visited Jiangsu sunshine group.

The so-called "other means of production" for the current sunshine group, the most typical is cloth.

In the sunshine group party secretary and chairman Chen Lifen view, data barrier is a major obstacle to the pformation of digital sunshine.

If the data between clothing and fabrics, departments and departments, and between domestic and export markets get through, then the role of sunshine group and even the entire textile industry will be immeasurable.

Chen Lifen, chairman of Jiangsu Sunshine Group

It is worth mentioning that the term "immeasurable" has been stressed three times by Chen Lifen. It can be seen that data have a new meaning for sunshine.

At present, most of the textile enterprises in China belong to small and medium-sized enterprises. Most of these enterprises lack full application of information technology and Internet talents. Data island phenomenon is serious. The linkage efficiency of each link in the industrial chain is low, resulting in overstock and delay in delivery.

With the rapid development of new technology, a small number of enterprises have created industry myths through the unique big data system and corresponding process changes. There are also large scale customized production processes, making the value of the data enlarged. It is becoming more urgent to optimize the industrial chain and forecast the market demand and reasonable pricing of textile products in the future.

The sun is also groping ahead in the flood of data revolution.

At the research site, Chen Lifen presented Dr. Wang Jian (and Xuelang number) three points of pain that the sun needs to solve at present. In fact, it also represents the common problems of most textile and garment enterprises.

How to reduce or even eliminate sample proofing?

Two how to reduce inventory?

Three how to achieve customized production in garment industry?

Digital pformation from fabric to hand

The fabric industry itself is a pattern based industry, which makes artificial intelligence based image recognition technology promising in this field.

When it was first established in 1986, the sun group is still a nameless town run wool mill (formerly known as Jiangyin fine wool mill). Now the sunshine of the year has long been a famous wool textile production enterprise and high-end garment production base in the world. Its products are 50% domestic market and 50% overseas market.

But with the rising cost of raw materials and the disappearance of demographic dividend, technological pformation has become the only choice for these enterprises to break through innovation.

The digital pformation of sunshine group can be said to start with fabrics.

In the textile industry, after ordering, it is necessary to first sample, first design the process of the sample, then collect the raw materials, and then produce the products according to the technological process, usually after dyeing, re combing, spinning, weaving, repairing, dyeing and finishing, and finished product inspection.

So proofing is actually a process of "setting standards". Once the sample is confirmed, the whole process and every process will be solidified.

According to Chen Lifen, the average cost of a proofing is about two thousand yuan.

The problem is that proofing is not done once, and customers sometimes need to confirm it again and again.

From wool dyeing to finishing, one proofing takes about a month, and two times is two months. In this year, about 6000 proofing is done, and the cost is about 12 million.

In addition to direct costs, proofing is also a "hidden cost" that disrupts the efficiency of the entire assembly line process.

"If you can" figure out a map ", you can find the basic data and file data of the cloth before, and even if it is similar data, it can be called directly, so that the delivery time can be shortened for at least one month.

Chen Lifen referred to the "map to find the map", in fact, is to take the sample pictures given by the user, through the artificial intelligence image recognition function to find flowers, flowers and materials with the highest similarity in the warehouse.

Such seemingly simple behavior, without technical support, will be too heavy to imagine.

Because in the information warehouse of sunshine group, there are about more than 80 thousand species of fabric materials lying on the floor. Finding similar cloth from it is like looking for a needle in a haystack.

"So customers are constantly coming to the market, and the company needs to keep sampling. Two times, three samples are wasted.

But in fact, there are ready-made in the warehouse, sometimes the guest wants 300 meters of cloth, the storehouse has 300 meters, does not need to do at all.

Chen Lifen also talked about an incident that had just touched her a few days ago.

A brand dealer needs a special fabric. Chen Lifen remembers that there are many in the warehouse.

But when she got to the warehouse, she was dumbfounded. Everyone said that the material was indeed available, but it was impossible to find it, because the system did not get through. The fabric warehouse and the clothing warehouse system did not get through, and the domestic sales warehouse did not get through with the export warehouse, and the code was not shared, which made it very difficult to find.

"Although there are two dimensional codes on each package, only after scanning is it known.

And clothing production involves many links. Sometimes cloth stores sometimes exist in clothing stores, sometimes in fabric stores, or in some intermediate links in the workshop.

This should be a common pain point in the industry. "

Chen Lifen said helplessly.

Although it was discovered by memory later, it also made Chen Lifen determined to dredge up the sunshine data and build a big data processing system similar to the ET industrial brain.

In this system, there are not only basic data and archival data of cloth, but also information about production process data, recipes, raw material specifications and so on. At the same time, integrating upstream and downstream enterprises, such as clothing companies, materials companies, etc., has become a shared service platform for an industry.

For these more than 80 thousand varieties of fabric, Chen Lifen intends to put FRID chip labels on them in the future.

In this way, when you look for a picture again, you just need to compare the photos on the computer, and through the image recognition technology of artificial intelligence, the background will show where and where the fabric is produced.

"100% of the best similarity, or 90%, 80%, even if sixty or seventy of the similarity, can also" crack "from the system to tune away, direct use.

The immediate effect is to clear up the stock.

According to Chen Lifen estimates, more than 80 thousand varieties of fabric inventory is less than 3 million meters, if the price of 100 yuan per meter to calculate, that is 300 million yuan!

"If this system is used in other enterprises, or even the whole

The benefits it brings is immeasurable.

Dr. Wang Jian said.

Dr. Wang Jian, honorary mayor of Xuelang Town, chairman of Alibaba Technical Committee, visited sunshine group.

Another typical application scenario of AI image recognition technology is finding flaws.

In the domestic textile industry, the original grey fabric and the final product defect detection are still in the manual detection stage.

Manual detection often has many shortcomings, such as slow speed, high detection rate and poor continuity. Therefore, replacing people with machines to find defects becomes the common demand of the industry.

In the sunlight, the work of finding defects requires 2 people to complete.

One is responsible for finding, another is responsible for marking, recording data.

There are three ways to detect defects: billet inspection, intermediate inspection and final inspection.

After the inspection, a supplementary repair is necessary.

Among them, the defect of billet inspection is the largest, almost half of it can be filtered.

Chen Lifen estimates that according to the current tens of thousands of kilometers of fabric production in the sunshine, there will be nearly 1 million defects. To accomplish so many quality inspections, the current use of sunlight is almost 7 or 80 manpower.

"If a machine is used to replace a person, and the image recognition technology of artificial intelligence is used to find defects, two people save at least one person, leaving a person marking.

That is to say, save 50% of manpower.

Under the big cargo scenario, how to achieve small batch customized production?

Before visiting the manufacturing enterprises, titanium media found that enterprises with highly customized products generally encountered a common problem: what are the good solutions for products with lots of orders and small batch production? This is also a difficult problem encountered by sunshine group.

In the sunshine, the customized production of large goods has always been a barrier.

The so-called big goods, that is, the relatively large volume of products, these customers are relatively stable, the contract is signed for 35 years, but the characteristics are lots of order batches and short delivery time, which has caused great distress to the original production process, and on the other hand, it also caused the asymmetry of R & D information.

Sunshine customers include HNA, China Eastern, Air China, Xiamen Airlines and other airlines.

These airlines will make their orders almost once a week.

Take East China Airlines, orders for many times a month, a set of orders less than 3-5 sets, more than a dozen sets, and airlines because of the large flow of personnel, flight attendants training for up to two months on the job, which means that 2 months must be delivered, which led to the pipeline on the East China Airlines is a list, and later is HNA's list, the entire production scheduling, scheduling, scheduling and so on, we need to constantly adjust, not only inefficient, delay in delivery is also common.

But there is also one advantage of these big goods, that is, the styles remain unchanged for a long time, and basically remain in the three regular numbers of S, M and L.

Chen Lifen said that if the data can be used to analyze the quantity and frequency of orders per year, the number of garments with different specifications, and the most frequently used styles, sunlight can be prepared in advance.

In the case of stable data, sometimes even one year's inventory can be prepared.

"This order can be reduced to once a month or even once a year, and the efficiency of production lines will also increase significantly.

Of course, if the data shows that the number of shipments is decreasing and the order batch is decreasing, we will pay attention to whether the other person wants to change the payment.

For example, Hainan Airlines, historical data show that three years to replace it, then we can go to third years to automatically reduce the stock.

Another point of pain in customized production is professional wear.

Professional wear is also the main product of sunshine.

Since 2000, the sun has launched a series of brands such as Venetian's tailored clothing and Pompeii professional wear, among which the annual sales of Pompeii are over 1 million 500 thousand.

Compared with ordinary clothing, the biggest feature of professional wear is that the size of each person is quite different from body size. This makes the requirements of volume, board and tailoring extremely harsh.

In particular, the size of the volume, the details of the control is very strict, from the pocket width, to the precision of the neckline sewing, data has accumulated hundreds of thousands of pieces.

Such fine volume data means that cutting must also be very precise.

In the sunlight, the cutting is done in two times -- wool scissors and fine scissors.

For example, now the production of 10 thousand sets of clothing, age from 18 to 60 years old, then only men's clothing needs 60 models, women's clothing is also, there are many other special features.

The men's dress adds up to almost more than 120 samples.

When cutting, the similar specifications need to be filed, and each file will be completed through two tailoring.

For hair scissors, keep 1 centimeters of -0.5 centimeters with the template, and then fine cut.

Such a set will probably cut off 5 centimeters.

。

"If a year is calculated according to the output of 2 million 500 thousand sets, saving a centimeter is 25000 meters, that is 1 meters 100 yuan that is the cost of 2 million 500 thousand.

5 centimeters down is 12 million 500 thousand loss. "

Chen Lifen said.

In addition to fabric loss, manpower loss can not be ignored.

There are three main points in this respect: one is skilled manual labor, the second is slow speed, and the other is manual cutting.

Chen Lifen said that the current volume data in the sunshine has not been well analyzed and processed. In the future, we hope that through the integration and mining of these data, we can form standardized typesetting specifications, and automatically tailor the technology to achieve one-step precise tailoring, which will greatly reduce fabric loss and save manpower costs.

"So this is the typical problem of using data resources to exchange other resources."

Dr. Wang Jian said.

From the point of view of resources, if data link can not reflect the efficiency of resource utilization of the entire enterprise, digitization becomes a matter of informatization for informatization.

Chen Lifen hopes that the digital innovation of sunlight can precipitate data and technology to form a textile brain, which can provide services for the whole industry and enhance the competitiveness of the whole industry.

Using artificial intelligence to "awaken" the textile industry

In June 30th, at the 2018 snow wave conference in Wuxi, Jiangsu Wuxi Economic Development Zone (Taihu new town) and Ali Yun jointly announced the start of 2018 snow wave manufacturing AI challenge, focusing on fabric defect intelligent recognition, and developing the application of big data and artificial intelligence technology in fabric defect identification, to help improve the quality of industrial manufacturing.

Pain in fabric defect detection

The textile industry has been playing a pivotal role in our national economy. In 2016, the output of cloth in China exceeded 70 billion meters, and the output has been on the rise.

If artificial intelligence and computer vision technology can be applied to the textile industry, the value of the textile industry will no doubt be enormous.

Fabric defect detection is textile.

An important part of production and quality management, however, fabric defect detection has been done by human eyes.

Manual testing is slow and labor intensive, and is affected by subjective factors. There is a lack of consistency. This method seriously reduces the automation level of textile production process.

It is understood that manual detection speed is generally 15-20 M / min, at this speed, a single inspector can only complete 0.8-1 m wide detection, so cloth inspection and finishing link has become a bottleneck in the entire production process.

Manual inspection also has the disadvantage of relying too much on the experience of inspecting workers, and frequently trips detection errors and missed inspections.

On the stage of the main forum of Xuelang conference, Chen Lifen, chairman of Jiangsu sunshine group, also shared the necessity of applying AI in this field.

She mentioned that in the sun, the work of finding defects usually takes 2 people to complete.

One is responsible for finding, another is responsible for marking, recording data.

There are three ways to detect defects: billet inspection, intermediate inspection and final inspection.

After the inspection, a supplementary repair is necessary.

Among them, the defect of billet inspection is the largest, almost half of it can be filtered.

Chen Lifen estimates that according to the current tens of thousands of kilometers of fabric production in the sunshine, there will be nearly 1 million defects. To accomplish so many quality inspections, the current use of sunlight is almost 7 or 80 manpower.

"If a machine is used to replace a person, and the image recognition technology of artificial intelligence is used to find defects, two people save at least one person, leaving a person marking.

That is to say, save 50% of manpower.

Data are new means of production.

The AI challenge is Ali Yun Tianchi, following aviation, power and industry, and is another industry based AI competition.

Jiangsu Wuxi Economic Development Zone (Taihu new town) will rely on Ali Yun Tianchi competition platform to collect the best algorithm for intelligent recognition of fabric defects, and Jiangsu Sunshine Group has provided thousands of samples with precise annotation.

Together with ALI Tianchi, the AI challenge is another exploration of Jiangsu Sunshine Group's application of high and new technology to the field of prevention and control.

Jiangsu Sunshine Group has been a pacesetter in China's textile industry.

The leading position has participated in the design and production of the new concierge dress, including the 70th anniversary parades of the war of resistance against Japan, the honor guard of the PLA, and the production of the "Shenzhou eleven" astronauts' autumn winter clothing.

Jiangsu Sunshine Group also attaches great importance to the accumulation and application of data, and has worked with aliyun team to build a big data processing system based on ET industrial brain, and has completed a series of innovations based on this.

From the perspective of resources, data is a new means of production for sunshine (Group), and its ultimate value is to maximize the value of other means of production.

Dr. Wang Jian, the chairman of the technical committee of Alibaba group and honorary mayor of Xuelang Town, expressed his visit to Jiangsu sunshine group.

As the data provider of this competition, Jiangsu Sunshine Group provides abundant and perfect cloth samples, including cloth sample, sampling environment, defect judgement standard, and professional guidance of process experts, providing support from many aspects of software and hardware environment.

Converging 200 thousand scientists, Ali cloud helps create "textile brain"

The data of this cloth defect test contest covers all kinds of important defects in plain cloth of textile industry.

The data consist of 2 parts: the original picture and the annotated data of the defect.

These data will pform new productivity in about 200000 of the world's top scientists gathered in Tianchi.

The Ali cloud "Tianchi", which hosts the contest, is the largest intellectual platform in the world, bringing together about 200000 AI algorithm scientists from around the world.

Ali cloud provides the machine learning PAI platform for the team, and the team can apply for it.

In the final competition team plan, we must include deep learning as the main algorithm.