LPR-LRtoHR-OCR

Aim:

• Detection on vehicle license plate from low resolution traffic images.(Using yoloV5)
• Extraction of license plates from the images and zoom them 5x times. (Using openCV)
• Enhance the resolution of the extracted images. (Using Real esrGAN)
• Extract the license plate text. (Using pytesseract)

Dependencies:

• pandas
• numpy
• yoloV5
• pytorch
• openCV
• Real ESRGAN
• pytesseract

Workflow diagram

License plate detection output

Extraction of license plate

Low resolution output after zooming

Low resolution to High resoltution

Text extraction from license plate

Model Performance

• Mean Average Precision(mAP)=92.6%
• Precision=94.8%
• Recall=87.8%