In recent years, China has accelerated the implementation of garbage classification system. In May 2021, the National Development and Reform Commission and the Ministry of Housing and Urban-Rural Development jointly issued the “14th Five-year plan” for Urban Household garbage Classification and Treatment Facilities Development Plan, requiring that by the end of 2025, In the Beijing-Tianjin-Hebei region and its surrounding areas, in the Yangtze River Delta, in the Guangdong-Hong Kong-Macao Greater Bay Area, in the Yangtze River Economic Belt, in the Yellow River Basin, and in counties where conditions permit in the pilot ecological civilization zone, a household waste sorting and treatment system has been basically completed. Encourage other areas to actively improve the coverage of waste separation and treatment facilities.

Since then, garbage sorting has become a national policy.

You might think it’s dirty, backbreaking, hard work, and you might be right. But what you may not know is that there is also huge economic value in it.

China is a country with a large population and a large garbage generation country. In 2021, the domestic garbage output of China was 279.972 million tons.

Due to different statistical caliber, the market value evaluation of waste classification market varies greatly, but even according to conservative calculation, it is still a market of hundreds of billions. According to our investigation, the annual recycling value of ningbo city alone is as high as 56 billion yuan.

If there is a method that can not only solve the pain, dirt and fatigue of garbage classification, but also improve the sorting efficiency and industry efficiency, and make “turning waste into treasure” a relatively easy job, is it a satisfactory choice expected by the society?

To that end, AI came.

1 Focus on Ningbo

On June 24, the World Bank’s executive board approved a loan of us $430 million to China, of which US $150 million will be used to separate waste in Ningbo.

Ningbo, as an eastern coastal city, has begun to take advanced waste management, recycling and prevention measures, which is in a leading position in China.

As a result, the world bank funds, mainly used in ningbo all category intelligent recycling bins fall to the ground, the project management combined with wisdom, block chain accounting technical advantages such as garbage classification user guide for delivering, the recycling process for big data analysis and management, the wisdom, for the living garbage classification side + recycling the pickup recovery control management, cost optimization, data analysis, This also marks ningbo is about to enter garbage classification “number wisdom” management era.

Opportunities are only for those who are prepared, and Ningbo Weilan is such an enterprise. Previously, it was a manufacturer of power distribution cabinets for the Internet of Things. Fan Shijie, the head of the company, paid close attention to market dynamics. At the end of 2018, when he saw THAT CCTV began to promote waste sorting, he immediately had an idea — waste sorting cabinets are essentially not very different from power distribution cabinets for the Internet of Things.

Fan Shijie immediately began to investigate the market, believing that the waste sorting cabinet produced in the market at that time could be made with Weilan’s technology, so he immediately put into research and development in November 2018, and produced a prototype in March 2019.

But in practice, he also discovered that garbage sorting cabinets actually involve a complex economic loop. After the waste glass, for example, only 70-80 yuan per ton, and the paper board per ton is as high as 2000 yuan of above, but waste paper board need detailed classification, because of the different types of recycling the price also is different, so if the equipment can be on the front end to more accurate classification of garbage, economic benefits and the recycling loop has much more room for optimization.

Of all the rubbish, the most difficult to deal with is kitchen waste. It may be hard for the average person to imagine that smelly, perishable kitchen waste actually has high economic value, but it is also the most difficult to dispose of kitchen waste. A common problem is that residents are not careful enough in sorting, mixing other waste with kitchen waste, which involves two actions — further sorting and alerting residents to similar problems.



So the real challenge is to increase the sorting of kitchen waste, because it’s a lot harder to identify. Strictly speaking, the first two generations of products without automatic bag breaking + kitchen waste identification can only be called “automatic classification cabinet”. Only when these two functions are complete can they be called “intelligent classification cabinet”.

Fortunately for Fan shijie, the company recruited AI technology director Du Yipin. We all know there is a shortage of AI talent, even fewer people with full AI development experience.

Before Du Yipin came to Weilan, he also knew that the company did not have a complete AI team, but he had confidence because he had mastered the new tool — Baidu EasyDL.

The common philosophy of EasyDL (and AutoML/DL) is to “use AI technology to help people design AI”. In other words, in the process of developing an AI model, data processing, feature extraction, model selection, parameter adjustment, training and deployment are all highly automated by utilizing AI capabilities. After a simple study can be used.

Du Yipin with EasyDL easily completed the kitchen waste identification model, but a problem, then need for kitchen waste recognition higher AI force, if returned to cloud computing, requires companies to acquire large amount of resources such as servers, the GPU card, this greatly increased the profit itself like a thin blade of the cost of recycling industry.

At this time, the engineer gave him advice. Instead of buying huge AI computing equipment, he only needed a Baidu “EdgeBoard embedded AI computing card” and two ordinary cameras to solve the problem.

When Du Yipin to less than 2000 yuan a piece of EdgeBoard calculation convergence to the wei rings in the garbage sorting cabinet, he said something very kindness: “our equipment costs increased by only about 2-3%, let us from the era of automation entered the era of artificial intelligence, have to say that this technology is the industry really need landing.”

In fact, the value of baidu fly blade is the precast it according to the actual demand of flexible collocation solution, even ningbo wei rings so that the company only a few know about the AI technology personnel of enterprises, can also use fly OARS, rich products and ecological from software to hardware, from training to the deployment, fix the whole equipment of AI, Enterprises only need very low cost, to bring the company to a new level of development.

Fan Shijie thus very feeling said, after upgrading to the third generation, Ningbo market, basically there is no our opponent, this is the power of AI.

In fact, Ningbo Weilan is just one of the rising army of waste recycling industry.

Garbage classification can make the utilization of garbage resources more profit, create more value for the industry and society, so it is necessary to carry out vigorously. From the perspective of industrial chain coordination, the front-end garbage classification service, the mid-end garbage classification and transportation, the terminal kitchen waste treatment, leachate disposal, garbage incineration and other related enterprises will benefit from this.

2 Is no longer just a “first mover”

In an enterprise, only one or a few people who know AI are taking the lead. For this phenomenon, Wu Tian, vice president of Baidu Group, calls it the “AI pioneer” stage.

As for AI landing, Wu Tian shared the practice path of landing observed through extensive cooperation with industrial partners. She described this path as three stages, that is, pioneer exploration stage, workshop application stage and industrial large-scale production stage.

Wu Tian found that the AI pioneers in enterprises are often not front-line business personnel or senior managers, but some senior engineers in the company. They understand the business, understand the scene, have ideas, want to break through the traditional business status quo, hope to use the power of AI out of a different way.

What we call the garbage sorting industry is such a typical scenario.

The huge market space of garbage classification has attracted investors and entrepreneurs, which has become a hot topic in the current venture capital industry. According to Qichumcha data, the number of garbage sorting related enterprises registered in China shows a trend of rapid increase. In 2016, the number of garbage sorting related enterprises registered was less than 1,000, and in 2020, the number was 3775.

However, it can be expected that traditional garbage sorting enterprises have obvious difficulties in recruiting AI development specialists of a certain level in the context of a large shortage of AI talents in the whole society.

For this situation, flying paddle EasyDL, BML and other low threshold enterprise AI development tools, greatly reduce the difficulty of garbage classification enterprise development, there is a “one person can support an enterprise” situation.

For example, the AI technology director of Ningbo Weilan told the author that before cooperating with EasyDL, we also investigated other companies, including Ali, Huawei and foreign platforms, and finally found that EasyDL is relatively better to use and more humanized, he said: “The great advantage of EasyDL is that it can be operated without advanced algorithmic foundations and deep learning knowledge. It’s easy to integrate into the system.”

3. Anti-fragmentation with ecological capacity

In fact, EasyDL offers a complete end-to-end solution, including EasyData for data services and EdgeBoard hardware for edge computing.

It has to be mentioned that EasyDL can still achieve good model effect under the condition of small amount of data, which has great practical significance for many enterprises in the stage of “stone-throwing”.

For example, Shenzhen Guigu Intelligent Co., LTD., which cooperated with Watsons in Hong Kong, China, received a project from Hong Kong Environmental Protection Bureau in 2019 to do a project to identify bottles. Although they look simple, bottles come in many different shapes. Guigu Intelligence used other open source code at the beginning, but the recognition rate was less than 60%, but after changing to EasyDL, the accuracy rate reached more than 90% in a short time.

Surprised to valley intelligent technical personnel, training in more than 100 bottles, 800 photographs, with EasyDL after training recognition rate and accuracy, barely meet the practical requirements, “this algorithm sounds easy, but we found that the algorithm can achieve similar effect on international company estimate is 2, 3, One of the more difficult tests was to swap the bar codes of two bottles of similar height, requiring them to be readable. In this experiment, only the EasyDL machine was able to identify it. None of the other companies’ equipment was able to identify it.”

This typically reflects the training effect of EasyDL in the case of small data volume.

In fact, from the above stories, we can easily find some commonalities, that is, although the scene is very fragmented, the actual practice is very difficult.

In the current wave of AI implementation, one controversial area is how to “de-fragment” excessively fragmented scenarios, whether to use more professionals + more costly customized solutions, or EasyDL, which has a lower learning curve and can be used by enterprise business personnel with a little training.

We conclude from these stories, let the enterprise developers themselves with AI ability is undoubtedly the best pieces of strategy, but it must be established in relying on the depth of the learning platform on the height of the industrialization, in the process, besides EasyDL, BML, the author also met just as electricity page of the choose and buy of EdgeBoard page, Baidu Feioar is equipped with a complete anti-debris capability by not missing every pain point and providing end-to-end one-stop solutions to the model that has been thoroughly honed through tens of thousands of scenarios and the data problems and software and hardware integration problems that may be encountered in the practice process.

Garbage sorting is just an epitome of the “digitalization” of our whole society. When AI begins to perform its magic in such a traditional field, it seems that we are closer to the goal of INTEGRATING AI into industrial production.

Baidu AI developer community ai.baidu.com/forum provides a platform for developers from all over the country to communicate, share and answer questions, so that developers can no longer “fight alone” on the way of research and development, and find better technical solutions through continuous communication and discussion. If you want to try all kinds of AI technologies and explore application scenarios, join baidu AI community immediately. All your ideas about AI can be realized here!

Scan the qr code below, add the little assistant wechat “JINGdong card, small customized peripheral, mysterious gift box, luggage” and more benefits you come to get ~