HelmetDetect is an application developed as a project for the Digital Image Processing (DIP) module at the National Institute of Business Management (NIBM). The primary objective of this project is to detect whether individuals wear helmets or not, aiding in enforcing safety measures, particularly in contexts such as road traffic management.
The images used for training were annotated using LabelImg to define classes, particularly distinguishing between individuals wearing helmets and those who are not.
For model training, we employed YOLOv5, a state-of-the-art deep learning algorithm for object detection. YOLOv5 offers a balance between accuracy and speed, making it suitable for real-time applications like ours.
To facilitate easy access and collaboration, we utilized Google Colab for model training. You can find our training notebook here.
A demonstration of the application in action is available via this video link.
We have deployed a web version of the application for user-friendly access. The source code for the web version is available in the web branch.
Feel free to explore the repository and contribute to the project! Your feedback and suggestions are highly appreciated.