PaddleOCR使用笔记-文字识别训练、评估、推理
PaddleOCR
Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)
项目地址:https://gitcode.com/gh_mirrors/pa/PaddleOCR
免费下载资源
·
PaddleOCR使用笔记
Linux环境下
PaddleOCR develop版本安装及使用(第二次尝试)(成功)
文字识别训练
在https://gitee.com/paddlepaddle/PaddleOCR/blob/develop/doc/doc_ch/models_list.md中下载英文识别模型的训练模型:
在https://gitee.com/paddlepaddle/PaddleOCR/blob/develop/doc/doc_ch/algorithm_overview.md下载rec_mv3_none_bilstm_ctc模型:
tar -xf ./pretrain_models/en_ppocr_mobile_v1.1_rec_train.tar -C ./pretrain_models/
tar -xf ./pretrain_models/rec_mv3_none_bilstm_ctc.tar -C ./pretrain_models/
export CUDA_VISIBLE_DEVICES=1
python3 tools/train.py -c configs/rec/multi_languages/rec_en_number_lite_train.yml 2>&1 | tee train_rec.log
python3 tools/train.py -c configs/rec/multi_languages/rec_en_number_lite_train2.yml 2>&1 | tee train_rec_aug.log
python3 tools/train.py -c configs/rec/multi_languages/rec_en_number_lite_train2.yml -o Global.checkpoints=./output/rec_aug_en_number_lite_crown/best_accuracy 2>&1 | tee train_rec_aug.log
文字识别评估
export CUDA_VISIBLE_DEVICES=2
# GPU 评估, Global.checkpoints 为待测权重
python3 tools/eval.py -c configs/rec/multi_languages/rec_en_number_lite_train.yml -o Global.checkpoints=output/rec_en_number_lite_crown/best_accuracy
export CUDA_VISIBLE_DEVICES=2
# GPU 评估, Global.checkpoints 为待测权重
python3 tools/eval.py -c configs/rec/multi_languages/rec_en_number_lite_train2.yml -o Global.checkpoints=output/rec_aug_en_number_lite_crown/best_accuracy
文字识别推理
python3 tools/infer_rec.py -c configs/rec/multi_languages/rec_en_number_lite_train.yml -o Global.checkpoints=output/rec_en_number_lite_crown/best_accuracy Global.infer_img=doc/imgs_crown_crop_test/
python3 tools/infer_rec.py -c configs/rec/multi_languages/rec_en_number_lite_train2.yml -o Global.checkpoints=output/rec_aug_en_number_lite_crown/best_accuracy Global.infer_img=doc/aug_rec_crown_test/
创作不易,喜欢的话加个关注点个赞,❤谢谢谢谢❤
GitHub 加速计划 / pa / PaddleOCR
41.53 K
7.59 K
下载
Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)
最近提交(Master分支:3 个月前 )
7bbda2bc
6 天前
1d4e7a80
8 天前
更多推荐
已为社区贡献15条内容
所有评论(0)