Back to Blog

Detecting Personal Protective Equipment (PPE) Using Machine Learning

Personal protective equipment detection through  machine learning

There are skyscrapers and architectural marvels all around us. Along with an ever-evolving development in healthcare and major advances in engineering. But we live our daily lives clueless about what it takes to reach this level of progress. With 7.6 million employees and $1.3 trillion in annual expenditure, construction is considered to be one of the largest sectors in the U.S. But it is also the most hazardous industry with a high number of workplace accidents and injuries.

The increasing number of casualties has made workplace safety one of the topmost priorities in many sectors including manufacturing, construction, healthcare, and logistics. The U.S. Occupational Safety and Health Administration (OSHA), expects industries to monitor the usage of the right Personal Protective Equipment (PPE) at the workplace. FDA regulates pharma industries to protect employees as well as products from any type of contamination. And with the Covid–19 pandemic, wearing a mask in public has become an essential safety requirement. 

Most workplace injuries can be easily avoided by just wearing the appropriate personal protective equipment such as helmets, gloves, safety vests, goggles, and safety boots. However, there is a constant check needed with automatic and real-time detection of non-compliant workers in using Personal Protective Equipment (PPE). Artificial Intelligence solutions using computer vision have the ability to solve this problem. Artificial Intelligence and machine learning can be an effective way to detect PPE violations and give alerts for any safety protocol breach. 

Here we discuss how AI and machine learning can be used for PPE detection to improve safety processes in many industries.

Why is Personal Protective Equipment (PPE) important?

The usage of PPE is important in many industries as it proves to be the last line of defense against injury or death. They are designed to protect workers from injuries due to physical, electrical, mechanical, or chemical contact. Establishments are required to do round the clock safety monitoring of entire sites to detect PPE violations.

What does work-related injury cost-1 Image source: Cortexica

Although wearing PPE is compulsory in workplaces with hazardous environments, workers may sometimes forget to wear them which may lead to an accident. The U.S. Bureau of Labor Statistics (BLS) report stated that the manufacturing industry reported 395,300 workplace injuries in 2019. According to another study, 98% of workers say they do not wear PPE when they are supposed to. For example, even though head injuries are the most common and fatal (constituting 9% of all injuries), 84% of these injuries are caused due to not wearing a helmet. Such injuries can easily be prevented with a system that can constantly check and monitor for PPE compliance.

hard-hat-and-vest

Usually, the responsibility lies with supervisors to check whether every worker is following the guidelines of wearing PPE, but this method may sometimes be ineffective and costly. Machine learning and computer vision can automate this process of detecting whether the workers are wearing the required PPE.

 

Overview of AI-assisted personal protective equipment detection using computer vision

Advancements in machine learning and computer vision have given rise to novel applications that can minimize latency, process data in real-time, and provide compliance monitoring of PPE. These systems can easily spot workers who are not wearing, or are incorrectly using PPE. Moreover they provide innovative solutions for audits and for generating reports automatically. 

What an AI-assisted PPE detection system can offer:

  • Real-time analysis using advanced machine learning algorithms to ensure compliance on wearing the right PPE according to the working environment.
  • Complete detection using image recognition, on the face, body parts, and person.
  • The system can be integrated with existing CCTV cameras, which can monitor designated areas, identify PPEs, and give alerts in the absence of required PPE.

PPE - Workplace safety

How does the Skyl computer vision model automate workplace safety and surveillance?

Inspecting and monitoring the proper use of PPE can be automated using Skyl.ai’s personal protective equipment detection model. Using CCTV cameras integrated with Artificial Intelligence, an object detection model can identify if the workers are wearing the right PPE. Skyl’s AI-assisted PPE detection using image recognition can automate workplace surveillance and monitoring. 

Try the inference API

skyl-model-on-PPE-detection

Try PPE Detection Model

Skyl.ai’s computer vision model automates PPE surveillance and inspection, detecting the lack of PPE to ensure safety. The machine learning model uses state-of-the-art neural network technology to identify different kinds of PPE. The input for this machine learning model is various images of workers wearing a combination of PPE, along with images where the workers are not wearing any protective equipment. The image data is labeled using Skyl Labelwise. For example, to detect a helmet and gloves on a worker, bonding boxes would be used to differentiate between them and name them. This task is done for the entire set of images. 

The model is then trained with a small feature set with equally distributed labeled categories. After deployment and implementation, it was found that the machine learning model could identify different kinds of PPE with 96% accuracy. 

This model is then integrated with a CCTV camera and parameters are set to alert the supervisors if a certain PPE is missing on the person. 

Skyl.ai’s machine learning model is able to automate PPE detection and monitoring successfully. The cutting-edge technology enables safety officers to ensure that everybody is wearing the required PPE on-premise. The model ensures that workplace accidents are reduced by 90%, decreasing employer liability costs and improving efficiency.

    

Comments