extract information values in id card, citizen identity card, passport, ...
Detect 4 corners of the id card, passport,… then crop the image. That makes the word detection task easier (reduce noise, variances, …) In this task, using SSD MobileNet v2 to detect corners. SSD has fast inference speed, MobileNet v2 uses low memory, operations but still has good accuracy
Detect some information likely id, name, birthday, … to cropping then use that to extract information task. In this task, using SSD like detect corner task but using Resnet FPN for feature extractor
Extract, get word from the image was cropped from word detection task. Using vietOCR to extract information on images. VietOCR is a pre-trained Vietnamese model that was trained on 10m images. GitHub - pbcquoc/vietocr: Transformer OCR
Clone the project
git clone https://github.com/whoisltd/OCR-ID-Card.git
Go to the project directory
cd OCR-ID-Card
Install packages
pip install -r requirements.txt
Start the server
python3 main.py