pengbohua / deepocrzh

Deep OCR for recent project

中文光学字符辨识(OCR)

步骤

安装需要的环境: Python 3.7.13 CUDA 11.x

pip install -r requirement.txt

detectron2 (推荐)

!pip install layoutparser torchvision && pip install "git+https://github.com/facebookresearch/[email protected]#egg=detectron2"

efficientnet

!pip install "layoutparser[effdet]"

paddle

!pip install "layoutparser[paddledetection]"

使用detectron2可能存在的兼容问题

环境中包含tools, 解决paddleocr安装包兼容问题(PaddlePaddle/PaddleOCR#1024) 5.31 Efficient和paddle环境不存在上述问题

将paddle中.tools替换成paddleocr.tools

mv replace_paddleocr.py pathToYourPaddleOCR/paddleocr.py

colab

mv replace_paddleocr.py /usr/local/lib/python3.7/dist-packages/paddleocr/paddleocr.py

Run the code

先运行extract_image.py将pdf中的每一页存成图片

python extract_image.py

Layout Analysis

python layout_analysis.py

---------------------------- 文字信息抽取 -------------------------- 之后下载模型

python3 image2text.py

可视化(Optional)

python3 visualization.py