DriveAssist

A system that uses computer vision to alert the driver when it detects traffic lights or road signs, helping them stay attentive and drive safely.

Inspiration

I wanted to create something with machine learning that would be genuinely useful in my day-to-day life. One of my relatives, who always recommends good project ideas to me, suggested that I build a system that alerts the driver when it detects road signs. I loved this idea and expanded upon it by also including traffic lights.

I will continue expanding on this project by adding support for new types of road signs in the future.

Features

• YOLO model for detecting traffic lights and road signs
• EasyOCR integration to extract speed values from speed limit signs
• Flexible input options: supports images, video file, and live camera feed
• Pygame integration for real-time audio output
• Custom logic for reducing false positive predictions from the model

Tech Stack

Languages: Python

Libraries & Tools: Ultralytics (YOLO), EasyOCR, OpenCV, Pygame, NumPy

Demo

https://drive.google.com/file/d/1akh2-6TfXNgXhZpitiBZOvNwW9qbtRs7/view?usp=drive_link

Deployment

1. Download the trained model:

   https://drive.google.com/file/d/1O77gxL7PurmNavvM7wI9FtN3UyhwLdOh/view?usp=drive_link

2. Fork this repository.

3. Clone the repository:

   git clone https://github.com/<your_username>/DriveAssist

4. Navigate to the directory:

   cd DriveAssist

5. Create a virtual environment:

   python -m venv <env_name>

6. Activate the virtual environment:

   Linux and macOS:

   source <env_name>/bin/activate

   Windows:

   <env_name>\Scripts\activate

7. Install dependencies

   pip install -r requirements.txt

8. Run inference:

   • Single Image:

     python main.py --model <path/to/model> --source <path/to/image>

   • Image Folder:

     python main.py --model <path/to/model> --source <path/to/folder>

   • Video File:

     python main.py --model <path/to/model> --source <path/to/video/file>

   • Camera Feed:

     You need the index of the connected USB camera (e.g. usb0).

     python main.py --model <path/to/model> --source <index_of_usb_camera>

Train Your Own YOLO Model

To train your own model, you can use the following notebook: https://colab.research.google.com/drive/1lJACvS_HwbCpCmPS5d_K2_c1bcA_c3JR?usp=drive_link

Dataset

To train the YOLO model, I created my own dataset, consisting of 370 dashcam images, and annotated it using the Roboflow platform. The dataset can be found here: https://drive.google.com/drive/folders/1fBAwNc9KgO--8yRwXxo6NV8RZowYXSSc?usp=sharing

Acknowledgements

• How to train a YOLO model - Edje Electronics