An open source VAT invoice OCR identification project

This issue recommends an open source VAT invoice OCR identification project——invoice。

Invoice is an OpencV-based VAT invoice OCR identification project that uses flask microservice architecture. The invoice types that can be identified include: VAT electronic ordinary invoice, VAT ordinary invoice, VAT special invoice; The identifiable fields are: invoice code, invoice number, invoice date, verification code, after-tax amount, etc.

environment

• python3.5/3.6
• Dependency installation:

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

• If there is a GPU environment, you can modify the installation requirements.txt version of the tensorflow-gpu, config.py file to control the GPU switch

Model architecture

YOLOv3 + CRNN + CTC

model

Model download address：https://pan.baidu.com/s/1bjtd3ueiUj3rt16p2_YQ2w

Place the downloaded model folder models under the project root directory

Service initiation

python3 app.py

The port can be modified by itself, and the service call address：http://…: [port number]/invoice-ocr，example：http://127.0.0.1:11111/invoice-ocr

example

The test tool uses postman

• VAT e-ticket

• VAT special general ticket

• VAT ordinary ticket