echizen321/alpr repository overview

⁠Usage

Run docker-compose up with the b/m docker-compose.yml and test endpoint with b/m API, body and check against expected results

⁠docker-compose.yml

version: '3.3'

services:
    alpr:
        ports:
            - '6891:6891'
        volumes:
            - '/home:/home'
        restart: unless-stopped
        environment:
            - LP_THRESHOLD=0.3
            - OCR_THRESHOLD=0.4
            - MIN_LP_LENGTH=3
        image: echizen321/alpr

⁠docker-compose parameters

1. LP_THRESHOLD (default 0.2)

lower bound for detection of license plate

1. OCR_THRESHOLD (default 0.4)

threshold for reading license plate text

1. MIN_LP_LENGTH (default 3)

minimum length of text read to be accepted as license plate

⁠API

⁠Endpoint and Body

POST
http://172.17.0.2:6891/ALPR
{
	"fileLocation": "/home/plato/Desktop/license_plate_recognition/cars_data/sg.jpg"
}

⁠Expected Output

{
    "cars_detected": true,
    "license_plates": [
        "SGL70IB"
    ]
}

⁠Workflow: how the code works

1. try to detect vehicle

2a) if vehicle exists, check the vehicle for LP (tighter constraint for LP detection)

2b) else, check original image for LP anyway (in case a cropped vehicle was not detected)

1. if have LP, check LP for contents (OCR)