INPR 0.1

A deep learning based project to detect indian number plates

Indian Number Plate Recognition

An open source Indian Number Plate Recogniton library built using deep learning.

For GUI version please go to this repo.

Version and updates

• v0.01

  • Detect number plates and extract the texts from it.

• Docker image comming soon!

Screenshots

Installation

This library can be installed on windows and linux just follow the below steps based on your OS.

for windows

1. I would suggest before installing please create an anaconda environment.

conda create -n "env-name" python==3.7

2. Install torch and detectron2

pip install torch==1.7.1+cpu torchvision==0.8.2+cpu torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

python -m pip install git+https://github.com/facebookresearch/detectron2.git

3. Then install INPR package

pip install INPR

for linux

1. install torch and detectron2

pip install torch==1.7.1+cpu torchvision==0.8.2+cpu torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
python -m pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.7/index.html

2. Then simply install the INPR package

pip install INPR

Usage

To detect the plates here is an example

    from INPR import detect_plates
    import matplotlib.pyplot as plt

    im = 'test_img/kia.jpg'
    grap,op,img = detect_plates(im)
    plt.imshow(grap)

In the above example the method "detect_plates" returns three values which are:-

1. the image along with detected plates
2. the bounding boxes of the detected plates
3. the passed image iteself

Optionally you can also try to extract the text present inside the detected plates by the following example:

    from INPR import fetch_details, detect_plates

    im = 'test_img/kia.jpg'

    grap,op,img = detect_plates(im)
    num_plate_text = fetch_details(op,img)
    
    print(num_plate_text)   

Feedback and queries

If you have any feedback, bugs, queries please reach out to me on Linkedin and also raise an GitHub issue about the same.