Automatic Number Plate Recognition with SSD model

Table of Content

Demo
Overview
Motivation
Technical Aspect
API Call
Deployment
Execution
Technology Used
Authors

Demo

Overview

This project is about locating the vehicle number plate and capturing the registration number from there using SSD object model and OCR technology.

Motivation

This model has several use cases like - granting access to vehicles based on registration number - capturing parking violations - maintaining a record of vehicles entering and leaving a parking lot

Technical Aspect

This project uses SSD object detection model.

API call (via postman)

  POST http://127.0.0.1/predict

Parameter	Type	Description
`image`	`string`	`base64 notation of image`

Deployment

To deploy this project, please follow the below steps in the same order

create environment

  conda create -n <envname> python=3.6 -y

activate the environment

  conda activate <envname>

install the requirements file

  pip install -r requirements.txt

download the .pb file from below location and keep it in datasets/experiment_ssd/2018_07_25_14-00/exported_model

    https://drive.google.com/drive/folders/1ZNUoxpuvylhEOAflXE6c_0JmaEuTnACF?usp=sharing

create a working directory to hold this project and use the below git commands to push work directory contents to your git repo

    git init
    git add . && git commit -m "first commit"
    git remote add origin https://github.com/.......git
    git branch -M main
    git push origin main

Execution Preparation and Steps:

Go to GCP platform
- Create a project like ANPR (for example)
- Go to Navigation Menu (left hand corner) >> API & Services > Dashboard
- Enable API & Services
- Go/Search for Cloud Vision API and Enable
- Left side Credentials > Top Create Credentials > Generate API Key and copy this
Open the project in pycharm
- Go to rest-server.py line 64 numberPlateVal = detect_license_plate(ik)
- Navigate to detect_license_plate method in getNumberPlateVals.py line 5 url, replace key= with the current key
- Start the service via python rest-server.py
Go to https://base64.guru/converter/encode/image and convert selected image to base64 format.
start Postman service
- POST with json format key "image" and value "base64 encoding of image" to URL http://127.0.0.1:5000/predict
- The output will be base64 string for the cropped portion of the image where the number plate is located and the value of number plate like "numberPlateVal": "KA01MR8041".
- This dict – key and value will come after passing thru the OCR
In real time when getting data from camera, we do cv2.videocapture and cam.read to get continuous image from the camera.