Automatic License Plate Reader Application

This repo contains an implementation of an application which is used for Automatic License Plate reading and offence booking.

Overview

Automatic License Plate Recognition (ALPR) is the technique of extracting the information from an image or a sequence of images of Vehicle's Number Plate. Considering the scenario of automation and the developments in artificial intelligence, machine learning and deep learning, everyone is trying to switch towards automation. The current system for issuing fines by the transport department is one such scenario where, making the whole process automatic would help save massive amount of time. The current model of issuing fines to violators is time consuming and requires a lot of human effort. The photos clicked by the cops are sent to the Regional Transport office for verification and the fines are issued considering the offence registered against the vehicle owner. A feasible way to make this process quicker and more reliable is by reducing the human intervention while carrying out each part of the procedure. The proposed web application for automatic license plate recognition aims at reducing the human interference by providing a solution for real time license plate detection and recognition. The details of the owner would be visible on the app and the offence for which the vehicle owner is being booked will be a manual choice. The receipt will be sent directly to the owner by email. It will reduce the chances of false imposition of fines. As it reduces human intervention, the chances of manipulating the amount to be paid by the owner are also less.

Method

App UI

For more details, refer to our project report: https://drive.google.com/file/d/1T2CuI0HmKRiUyxGygH9FH6TN2xa3mQjO/view?usp=sharing

Usage

Installation

git clone https://github.com/KapilM26/license-plate-reader.git
cd license-plate-reader
pip3 install -r requirements.txt

It is recommended to create a virtual environment for this project.

Preliminary application setup

Email.config

The application uses a file called email.config to get the Email ID and password of the account to send the email to the offender. Create the file email.config inside the alpr/ folder. The first line of this file should have the email ID and the second line should have the password. Make sure not to push this file to Github!

Download/train model

The pretrained YOLOv3 model can be downloaded from here: https://drive.google.com/file/d/14sR8yAX-ByUxUGARqr3YU5vmnuaNeSbZ/view?usp=sharing Download this file and put it inside alpr/models. Alternatively, you can refer to the train_yolo_alpr notebook to train a custom model.

YOLO implementation used: https://github.com/eriklindernoren/PyTorch-YOLOv3

Dataset used for training: https://web.inf.ufpr.br/vri/databases/ufpr-alpr/

Running the application

cd alpr
python3 manage.py runserver

This will run a local server on http://127.0.0.1:8000/ You can also run a server on a particular IP address by specifying it (along with the port) in the runserver command.

python3 manage.py runserver <IP>:<PORT>