top of page
  • Writer's pictureMAILab2021

MAILab Newsletter: Safety AI (February 5, 2024)

You can watch MAILab's QSS quarantine-security-safety integrated solution video for factories on YouTube.

Industrial safety has now gone beyond a social issue to become not only a political issue but also an economic issue. Until now, companies have poured all their investment and time into consumer satisfaction and user experience, but now competition between companies on employee satisfaction has become inevitable. After the pandemic, employees have come to realize that their lives and safety are more important than their salaries or welfare treatment, so we have come to a situation where we have no choice but to value the health and safety of employees rather than amicable labor-management relations in order to secure employment.


Now, the concept of ESG no longer makes sense to us. However, ESG of environment-social responsibility-governance is still too broad, and if corporate social responsibility and transparency are overemphasized, businessmen's political voice becomes too strong and, conversely, external influence becomes too strong. Interference in corporate management may intensify. What is more realistic and urgent for us who have experienced a historically unprecedented pandemic is EHS, that is, sustainability through E(environment)-H(health)-S(safety). Sustainability is very important whether for a country or a company, but it feels like the understanding is too biased to one side as ‘eco-friendliness = sustainability’. Health and safety are immediately appealing to both employees and customers, and are also essential elements for corporate sustainability.


Another reason why Korean companies have no choice but to invest in the health and safety of their employees is the Special Act on Serious Accidents. The official name is ‘Act on Punishment for Serious Accidents, etc.’ From the 27th, the Serious Accident Punishment Act also applies to workplaces with 5 to 49 employees.


Both pros and cons regarding the Special Act on Serious Accidents agree that the law alone is insufficient. This is because our country previously did not have strict laws on industrial accidents. For example, the Industrial Safety Act was also a very strict law. This is also why it is still questionable whether practical problems can be solved with stricter laws than before.


Artificial intelligence, especially Vision AI, is being discussed as a solution to industrial accidents. Many companies are already claiming that they are using artificial intelligence, especially vision AI, to improve industrial safety, but considering the overall level of vision AI technology in our country, we cannot help but feel that AI is only a pattern. In particular, it is claimed that AI analyzes the video captured by CCTV, which is just an RGB camera, and informs the level of risk, but it is questionable whether it can truly perform a preventive function. Above all, the special function of deep learning artificial intelligence is proactive rather than reactive tracking.


To put it more explicitly, expressing it as ‘boxing’ cannot serve as evidence that the preventive function for workplace accidents has been realized. ‘Boxing’, also called “bounding box”, is also abbreviated as ‘bbox’. It is a rectangle used to indicate the spatial location of an object in object detection, a part of vision AI, and is located in the upper left corner and lower right corner. Its location is determined by ‘boxing’ ultimately displays the location of the recognition target object.


Safety equipment (personal protective equipment: PPE), that is, safety helmets, safety vests, and safety shoes, as well as gloves and masks, may vary depending on the work site. When entering the workplace, it is necessary to check that safety equipment is worn one by one and continue to wear it during actual work. In this case, Vision AI can be used. Then, it is also important to identify who is wearing it and who is not, and then send a warning in real time. If it is detected in a specific place, it must be announced through a speaker, etc., which can cause problems not only in the atmosphere of the workplace but also in individual privacy. In other words, how to send an alert is as important as real-time detection. Additionally, in reality, it is extremely difficult to detect in a dynamic state and even recognize who the violator is.


When it comes to industrial accidents, it is just as important to provide first aid as soon as an accident occurs, just as it is to prevent the accident. This is because how quickly emergency measures are taken after an accident occurs and how quickly you can reach the hospital by ambulance are critical factors not only for your life but also for your recovery and health after the accident. It is very important to secure a golden time that can protect your life and health. For this purpose, a system is needed that detects accidents in real time using an algorithm trained with data in advance, makes an SOS call at the same time, and allows emergency measures to be taken quickly.


Of course, fires can also be quickly discovered and extinguished through Vision AI. So far, sensors have been used to detect smoke from flames, but if the ceiling is high and the area is large, smoke detection is not possible immediately. It is often detected only after the fire has spread to some extent. If a fire breaks out in a place such as a logistics center, because so many flammable materials have accumulated, the fire spreads quickly and people do not have time to evacuate, so it is likely to develop into a major casualty accident. Now, if a camera equipped with an AI chip detects flames in real time using the Vision AI function like the human eye, such large-scale fires can be extinguished early and casualties can be prevented.


Lastly, we cannot help but mention industrial accidents at construction sites. At construction sites, the situation is so serious that even self-deprecating voices say, ‘You have to be lucky to survive.’ More than 400 deaths occur every year, accounting for more than half of all industrial accident deaths. It mainly involves falls, collisions, and getting caught. The field that most desperately needs the Vision AI technology mentioned so far is the construction industry.


The role of Vision AI is similar in construction sites and general workplaces. It detects and warns in real time whether safety gear is worn, identifies the person not wearing it as necessary, and blocks access to hazardous areas, hazardous facilities, and operating equipment in advance through a virtual perimeter, and prevents fires as well as heavy equipment. It detects safety conditions and prevents accidents in advance, and if a disaster occurs, it quickly calls emergency personnel and takes emergency measures. The only characteristic is that the detection target area is large and varied by mixing indoors and outdoors, workplaces are distributed by floor, and the approach and movement of moving objects such as cars differ depending on individual zones inside. Simply put, construction sites are wide, diverse, and complex in terms of safety.

Despite the circumstances, safety issues at construction sites can never be given up, and I believe Vision AI will provide the solution. Currently, MAILab has not only developed Vision AI's hardware-software that addresses safety inside fulfillment centers and manufacturing plants, but is also researching artificial intelligence technology that can solve safety issues at construction sites. I must confess that the biggest problem is securing data. Ultimately, collaboration with construction companies is absolutely necessary.


Education cannot be overlooked when it comes to industrial accidents. If education has been simply based on past experiences, artificial intelligence technology based on big data, visualization through graphic analysis, and even prediction should be utilized. In particular, since many camera sensors are required to be based on Vision AI, there is no choice but to select and focus on areas with a high frequency of accidents. In other words, the analysis and prediction capabilities of artificial intelligence are essential, even if only because of cost.


Ultimately, we cannot help but worry about the relationship between cost and safety AI. In particular, with the development of vision AI and edge AI, investment in hardware, which is an edge device, is inevitable, resulting in cost issues. As briefly mentioned earlier, it may be possible to solve the problem with the wisdom of selection and concentration, but it is insufficient as a fundamental measure. The alternative to this is multi-function. For example, building a hardware-software-service that functions not only for safety but also security and quarantine. This is Maillab’s QSS (quarantine-security-safety) integrated AI solution. Technically, it has been upgraded from Narrow AI with simple functions to General AI with multi-functions.


Industrial accidents cannot be solved with laws, culture, or awareness. Now, artificial intelligence has already entered our lives as a means of solving this problem. The problem is that it is a cutting-edge technology and at the same time has high cost-effectiveness, so it is good to encourage companies and governments to actively utilize it. This too is a technical issue, and this extremely simple fact must be actively communicated. In addition, I think it is the company's historical responsibility to actively promote the fact that if we not only solve industrial accidents but also shorten the time to enter the era of artificial intelligence, we will kill two birds with one stone.



About the author Dr. Phil Yang: Chairman of CKT Group; CEO of MAILab; China Qingdao Municipal Government Economic Ambassador; Advisor, Tsinghua University Foundation, China; Adjunct Professor, Tsinghua University, China; International President of Jilin Vocational University, China; New York State University, Korea Branch, Chair Professor; Director, Konkuk University's Chinese Enterprise Research Institute


MAILab MAILab commercializes AI solutions for quarantine, security and safety with QSS integrated solution based on its own engine i2Brain, Vision AI Framework, MEGA Image AI Platform, and MEGA Industrial AI Platform through engineering and marketing organizations in the United States, Canada, China and India. MAILab offers Industrial AI such as QSS Integrated Solution, Medical AI such as congenital heart disease algorithm, Logistic AI such as Robotic AI Fulfillment Service, Beauty AI for sales and production of skin care and makeup products.


Contact Us:




bottom of page