Workplace Safety Identification
| Summary |
|
This project demonstrates the implementation of a Custom Vision ML model to the Vision AI DevKit to identify the presence of workplace safety equipment such as hardhats, vests, and safety glasses. As each object is detected, the information is stored and an alarm is triggered when a person that is not wearing the specified equipment is detected in the frame.
Partnered with Purdue University and Rockwell Automation, this model was a featured project in a scientific journal - read the published article here. |
| Implementation |
|
IoT Hub identifies the detected object data from the device and sends data to the cloud to further process using downstream services. Stream Analytics reads the data from the incoming event hub in IoT hub and parses the JSON and stores information into Azure SQL and Blob Storage. The incoming data set is deflated from JSON format to a structured format for further analysis. Azure SQL database stores detected object data (including time of detection) for generation of interpretive charts displayed on the website and Power BI reporting. The data is only kept for a 6 months to yearlong time frame. Blob Storage in Azure stores the same data for long term storage. Data older than one year or more is kept for auditing records and compliance. Web App Dashboard uses Azure SQL data to display the information on the webpage. |
 |
 |
| Software and Services used | Hardware | |
|
|
| Repository |
|
Find all relevant information, including code, pictures used for model training, and the Custom Vision model file for full implementation of this product here. Users are always encouraged to innovate and continue to improve the functionality of current projects. |
| Future Improvements and Project Suggestions |
|
This project has many different opportunities for improvement by other developers. The Custom Vision AI model could be further trained with additional images to identify equipment more accurately in a variety of scenarios or a system could be implemented to view images on the cloud before pushing specific pictures to train the Custom Vision model. Other opportunities could include: adding object tracking capabilities, enabling alerts to businesses or systems upon a safety violation, or adjusting the model or processing based on different industrial use cases.
Feel free to fork the project and contribute back any improvements or suggestions. Contributors and maintainers are encouraged. |
| About the Creator |
 |
Balamurugan Balakreshnan help customers by providing thought leadership in their Digital Transformation using Cloud, AI, ML, IoT, Block Chain.
You can learn more about what Bala is working on here. |
Leave a comment