CINE-Reader is an ML model developed to extract useful information from Moroccan identity cards, both new and old versions, with a 94% accuracy rate. Thanks to this model, filling out forms can be accomplished with just one click.
CINE-Reader is a combination of two models, namely a fine-tuned YOLOv8n and an EasyOCR model. Grayscaling and a bilateral filter are applied to the identity card image to facilitate the extraction of personal informations.
To run CINE-Reader on your laptop you should do the following steps:
pip install -r requirements.txt
cd ocr_project
python manage.py runserver
The process of extracting personal information from the identity card goes through several steps.
After detecting and cropping the regions of interest, a bilateral filter and grayscale transformation will be applied to the cropped images in order to reduce noise and enhance the images' features. This preprocessing step is essential for facilitating the text extraction process by improving the image quality and simplifying the color information.
The output of the program is as follows:
{ "first name": "ZAINEB", "last name": "EL ALAMI", "birth date": "05/12/1983", "CIN": "U1234567" }
To enhance the usability of our model, we developed a straightforward web application using Django, vanilla CSS, and vanilla JavaScript to facilitate this process. The web app consists primarily of two pages. The first page is dedicated to Identity Card deployment, while the second page displays the extracted data alongside the personal picture. Additionally, users have the ability to edit their information in case of inaccuracies.