This project uses Optical Character Recognition (OCR) technology to automatically extract text from receipts. The OCR functionality has been integrated into a local web app for easy access and usage.
The OCR models used in this project include a receipt corners detector based on YOLOv4, a line detection model and a CNN-LSTM based OCR model, all written in PyTorch.
- Clone this repository to your local machine
git clone https://github.com/vantuan5644/ReceiptOCR.git
- Navigate to the project directory
cd ReceiptOCR
- Install the necessary dependencies
pip install -r requirements.txt
- Start the local webserver
python server/app.py
- Access the web app by going to
localhost:5000in your web browser
The models used in this project were adapted from the following sources:
-
Line detection model: Zhou et al.’s 2017, EAST: An Efficient and Accurate Scene Text Detector.
-
OCR model: Quoc 2020, VietOCR
This project is licensed under the MIT License.