License plate detection and text extraction using both YOLO and OCR models

This is a license plate localizer using pre-trained YOLOv5, combined with a text extractor built with a pre-trained TrOCR.

The functioning is simple:

• The yolo model localize the plate(s)
• The ROI (Region Of Interest) is cropped
• The TrOCR model is used on the cropped image to extract the text from it

Usage

1. Install dependencies

Once you downloaded my repo, set it as your working directory and type in the python console:

pip install -r requirements.txt

2. Use your own images

In the line 25 you have to put the path of the image you want to submit to the models:

img = 'path/to/your/image'

Results

1. Example of an original image used

2. Cropping performed thanks to the Yolo model

3. Text extraction performed by the TrOCR model