Deep-learning-model-to-detect license plates/ Number plates of a car and read them-in-realtime using a custom trained yolov4 model and Tesseract-OCR (please read before executing)

Before starting with this,

MOVE the obj-license.data and custom cfg file into the cfg folder in the x64 folder in your darknet build.

MOVE the obj.names into the data folder in the x64 folder in your darknet build.

Also download Tesseract from here Link to download tesseract

Dont forget to download tesseract and change the path in the python file to the location of tesseract.exe in your local machine.

Working:-

Our yolov4 model is custom trained on license plates from the google open image dataset Dataset for yolov4 training

We use opencv to feed the video into the yolov4 model which feeds it to the custom trained yolov4 model which inturn gives us bounding boxes for the license plates.

Each bounding box is then cropped from the frame and passed on.

After some processing using opencv the images are passed onto the tesseract OCR which inturn outputs the license plate / number plate number

The output of the yolov4 detections will be saved in the demo folder in the x64 folder in your darknet build.