Vietnamese License Plate Recognition

This repository provides you with a detailed guide on how to training and build a Vietnamese License Plate detection and recognition system. This system can work on 2 types of license plate in Vietnam, 1 line plates and 2 lines plates.

Installation

bashCopy
  git clone https://github.com/Marsmallotr/License-Plate-Recognition.git
  cd License-Plate-Recognition

  # install dependencies using pip 
  pip install -r ./requirement.txt

• Pretrained model provided in ./model folder in this repo

• Download yolov5 (old version) from this link: yolov5 - google drive

• Copy yolov5 folder to project folder

Run License Plate Recognition

bashCopy
  # run inference on webcam (15-20fps if there is 1 license plate in scene)
  python webcam.py 

  # run inference on image
  python lp_image.py -i test_image/3.jpg

  # run LP_recognition.ipynb if you want to know how model work in each step

Result

Vietnamese Plate Dataset

This repo uses 2 sets of data for 2 stage of license plate recognition problem:

• License Plate Detection Dataset
• Character Detection Dataset

Thanks Mì Ai and winter2897 for sharing a part in this dataset.

Training

Training code for Yolov5:

Use code in ./training folder

bashCopy
  training/Plate_detection.ipynb     #for LP_Detection
  training/Letter_detection.ipynb    #for Letter_detection