一,安装easyocr:
1,官网:https://www.jaided.ai/项目代码地址:
https://github.com/JaidedAI/EasyOCR通过pip安装:
[root@blog ~]# pip3 install easyocr查看已安装的easyocr的版本:
[lhdop@blog img2]$ pip3 show easyocr WARNING: Ignoring invalid distribution -addlepaddle (/usr/local/soft/python3.7.15/lib/python3.7/site-packages) Name: easyocr Version: 1.6.2 Summary: End-to-End Multi-Lingual Optical Character Recognition (OCR) Solution Home-page: https://github.com/jaidedai/easyocr Author: Rakpong Kittinaradorn Author-email: [email protected] License: Apache License 2.0 Location: /usr/local/soft/python3.7.15/lib/python3.7/site-packages Requires: ninja, numpy, opencv-python-headless, Pillow, pyclipper, python-bidi, PyYAML, scikit-image, scipy, Shapely, torch, torchvision Required-by:
说明:刘宏缔的架构森林是一个专注架构的博客,地址:https://www.cnblogs.com/architectforest
对应的源码可以访问这里获取: https://github.com/liuhongdi/
或: https://gitee.com/liuhongdi
说明:作者:刘宏缔 邮箱: [email protected]
二,下载并安装模型
1,模型的官方地址:https://www.jaided.ai/easyocr/modelhub/2,官方站的下载很慢,我转存到了百度网盘,链接:
https://pan.baidu.com/s/13VsP9RbbTEguJtCl812XEA?pwd=lstn提取码:lstn 3,下载模型后安装:
- Linux:放到此目录下 ~/ .EasyOCR / model
- 例子:
[lhdop@blog model]$ pwd /home/lhdop/.EasyOCR/model [lhdop@blog model]$ ls chinese_sim.pth craft_mlt_25k.pth english_g2.pth zh_sim_g2.pth如果不手动下载安装model, 则第一次运行时会自动下载, 例:运行:
[lhdop@blog img2]$ python3 easy.py CUDA not available - defaulting to CPU. Note: This module is much faster with a GPU. Downloading detection model, please wait. This may take several minutes depending upon your network connection.
说明:这个时间非常漫长,建议手动下载
三,编写代码
easy.pyimport easyocr reader = easyocr.Reader(['ch_sim','en']) result = reader.readtext(r'/home/lhdop/img2/text.jpeg') print(result)
四,测试代码
原始图片: 返回结果:[lhdop@blog img2]$ python3 easy.py CUDA not available - defaulting to CPU. Note: This module is much faster with a GPU. [([[237, 41], [387, 41], [387, 127], [237, 127]], '推文', 0.9029050204465915),
([[36, 443], [761, 443], [761, 558], [36, 558]], '今天听到一个笑话:', 0.7776902087822264),
([[798, 448], [894, 448], [894, 552], [798, 552]], '美', 0.9923826635401589),
([[40, 562], [675, 562], [675, 674], [40, 674]], '国捅了欧盟一刀,', 0.6283286688234799),
([[708, 566], [894, 566], [894, 670], [708, 670]], '然后', 0.982831223026073),
([[30, 682], [479, 682], [479, 793], [30, 793]], '对着俄罗斯_', 0.9716240304123038),
([[543, 685], [933, 685], [933, 793], [543, 793]], '问"你怕了', 0.9087387627503959),
([[213, 817], [251, 817], [251, 853], [213, 853]], '99', 0.6924858343136447),
([[38, 812], [182, 812], [182, 908], [38, 908]], '吗?', 0.9797778345485638),
([[330, 800], [861, 800], [861, 915], [330, 915]], '欧盟也拿出刀_', 0.954007364936565),
([[28, 923], [895, 923], [895, 1034], [28, 1034]], '狠狠捅了自己-刀。也', 0.555011004631447),
([[30, 1039], [483, 1039], [483, 1153], [30, 1153]], '对着俄罗斯', 0.9807341978941745),
([[542, 1042], [937, 1042], [937, 1153], [542, 1153]], '问"你怕了', 0.7828333677200282),
([[38, 1172], [198, 1172], [198, 1270], [38, 1270]], '吗?', 0.9868630413885496),
([[211, 1177], [249, 1177], [249, 1223], [211, 1223]], '99', 0.8784055841595687),
([[330, 1161], [1017, 1161], [1017, 1273], [330, 1273]], '最后俄罗斯也拿出', 0.9547160164916341),
([[42, 1292], [250, 1292], [250, 1392], [42, 1392]], '了刀,', 0.8193539619850837),
([[290, 1282], [901, 1282], [901, 1394], [290, 1394]], '也捅了欧盟-刀', 0.6808263759334375),
([[596, 1418], [632, 1418], [632, 1450], [596, 1450]], '99', 0.5040398968561932),
([[45, 1403], [548, 1403], [548, 1514], [45, 1514]], '回答说"不怕!', 0.8787340291743182),
([[34, 1552], [247, 1552], [247, 1624], [34, 1624]], '翻译推文', 0.9337367415428162),
([[32, 1647], [605, 1647], [605, 1731], [32, 1731]], '22年10月128,14:31', 0.9352973260072716),
([[36, 1735], [528, 1735], [528, 1802], [36, 1802]], 'Twitter for iPhone', 0.8464620277583819)]
五,查看python版本:
[lhdop@blog img2]$ python3 --version Python 3.7.15
标签:1.6,15,easyocr,blog,https,lhdop,model,com From: https://www.cnblogs.com/architectforest/p/16847981.html