简要描述项目
2080ti单卡 cuda10.2 cudnn7.6(其实只要能跑paddlepaddle-gpu==2.0.2的环境就ok)
注意:git lfs 要求 git >= 1.8.2
Linux
curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.deb.sh | sudo bash
sudo apt-get install git-lfs
git lfs install
Mac
安装HomeBrew /usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
brew install git-lfs
git lfs install
下载后,将在data目录的A\B\训练测试集.zip,解压到data目录
-|data
|A
|B
|训练数据集
-|work
|PaddleOCR
-README.MD
-xxx.ipynb
A:在AI Studio上运行本项目 B:此处由项目作者进行撰写使用方式。
pip install -r requirements.txt
cd work/PaddleOCR && python3 tools/train.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0_2_cutout_suijipinjie_ksize=3_test.yml
python tools/export_model.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0_2_cutout_suijipinjie_ksize=3.yml -o Global.pretrained_model=./rec_chinese_lite_v2.0_1.0_cutout_suijipinjie_ksize3_test/best_accuracy Global.save_inference_dir=./inference/rec_inference_lite_cutout_suijipinjie_ksize3_test/
python tools/infer_rec.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0_2_cutout_suijipinjie_ksize=3.yml -o Global.infer_img="../data/A榜测试数据集/TestAImages" Global.pretrained_model="./output/rec_chinese_lite_v2.0_1.0_cutout_suijipinjie_ksize3_test/best_accuracy"