This project aims to detect number plate of a vehicle in a image or video and extract the number plate text using Optical Character Recognation(OCR). This system is implemented by using yolov5, EasyOCR and a Custom OCR model.
The dataset used to train yolov5 dectection :
Pakistan Vehicle Number Plate Dataset
Labeled licence plates dataset
Used Roboflow to annotate/label our data.
- Clone the repository
git clone https://github.com/ZohairAbbas/Number-Plate-Detection.git
- Change the working directory
cd Number-Plate-Detection
- Install dependencies
pip install -r requirements.txt
- Run the app
python -m uvicorn api:app --reload
The pipeline in the project is as follows:
- Train custom data using yolov5
- Custom number plate detection using yolov5 and saving plates as cropped images
- Apply the extracted plate to EasyOCR and Custom OCR
- Get plates-text in an array format
- Integeration of FastAPI