Real-Time-Bangladeshi-License-Plate-Detection-And-Recognition

Table of Contents

1. About The Project
   • Built With
2. Getting Started
   • Prerequisites
   • Steps
3. Result
4. Contact

About The Project

This project aims to develop a system which can detect Bangladeshi license plate (Bangla characters) and recognizes the characters in the license plate.The licenese plate information will be saved in a xls file along with some other informations such as location,time,cropped picture of the license plate and an unique id for each deteced license plate.

***Cite the dataset:

M. H. Tusar, M. T. Bhuiya, M. S. Hossain, A. Tabassum and R. Khan, "Real Time Bangla License Plate Recognition with Deep Learning Techniques," 2022 IEEE International Conference on Artificial Intelligence in Engineering and Technology (IICAIET), 2022, pp. 1-6.***

Built With

• Python

• Yolov5

• EasyOCR

• Raspberry Pi 4B (optional)

• Raspberry Pi Camerav2 (optional)

Getting Started

Before getting started, we need to have some data,software and access to some websites which is necessary for this project.

Prerequisites

1. Bangladeshi Bangla License Plate Dataset
2. Anconda or Pycharm for model training
3. Alernatively, we could also use Google Colab or similiar kind of cloud platform for model training
4. A webcam if raspberry pi with its cmaera is not used

Steps

• To train the model we will use google colab and run the whole project we will use anconda or terminal in case of raspberry pi.First,we need to create a virtual environment in anaconda.To create a virtual environment,open anaconda promt from start menu and enter following command

conda create -n {virtual environment name}

Activate the virtual environment by issuing:

conda activate {virtual environment name}

• We have created virtual environment.Now we need to prepare our data to train the model.We have already created a dataset of Bangladeshi license plate and it can be downloded from this link: Bangla License Plate.In all, there are 295 images with annotations in the dataset.And they are splitted into three parts;train(70%), valid(20%) and test(10%).Now to preapre the dataset for training we will use roboflow.They has a guidline on how to prepare a dataset for taining. We will use colab to train our yolov5 model.We could also use anaconda or pycharm to train the model in local machine.

• We trained the model in colab and downloaded the weights from colab.Two trained weights yolov5s & yolov5n will be included in this repository.

• Now make a directory in your preferable drive and create a folder where we will store all the necessery files.So,if needed, we change the directories in anaconda command prompt and create a folder and change directories to that folder.

mkdir <folder name>
cd <folder name>

• Now we clone yolov5 repository to this folder by issuing:

git clone https://github.com/ultralytics/yolov5 .

We need to delete the default detect.py that comes with yolov5 directory by using following command:

del detect.py

We will clone this repository for customized detect.py and trained weights.

git clone https://github.com/MahmudolHasan/Real-Time-Bangladeshi-License-Plate-Detection-And-Recognition .

We need to install and import python packages to run the detection by using:

pip install -r requirements.txt

we also need to install easyocr and xlswriter by issuing:

pip install easyocr==1.4.0

pip install xlsxwriter

pip install uuid

Now run following command to real time detect license plate:

python detect.py --weights yolov5s.pt --img 320 --conf 0.5 --source 0 --save-crop 

Note: use '--save-crop' argument to save the license plate information in an .xls file.

Result

Tested Cars pictures

The Output Result

Contact