目录
1. 安装依赖
2. 下载模型文件
3. 编写代码
4. 运行脚本
命令行模式
参考文献
一个Python的图像文本识别工具：easyocr.
这个模块支持70多种语言的即用型OCR，包括中文，英文，日文，韩文和泰文等。
注：OCR = Optical Character Recognition 光学字符识别
1. 安装依赖
pip install easyocr
2. 下载模型文件
由于联网下载模型可能会很长时间或下载出错，可以下载预训练的模型文件到本地。
下载链接：
Release Model files · JaidedAI/EasyOCR · GitHubhttps://github.com/JaidedAI/EasyOCR/releases/tag/pre-v1.1.6
文件说明：
| 模型文件 | 识别文字 |
|---|---|
| craft_mlt_25k.zip | 文字检测模型（CRAFT）（必须） |
| chinese_sim.zip | 中文（简体）模型（识别中文必须） |
| chinese.zip | 中文（繁体）模型 |
| latin.zip | 拉丁文模型 |
| japanese.zip | 日文模型 |
| korean.zip | 韩文模型 |
| thai.zip | 泰文模型 |
| arabic.zip | 阿拉伯文模型 |
模型文件放置位置：
-
Windows：C:Users用户名.EasyOCRmodel
-
Linux：~/.EasyOCR/model
例如：Windows路径如下：（Snow-Angel为我的计算机用户名）
3. 编写代码
文件名为 easyocr_main.py
import easyocr
reader = easyocr.Reader(['ch_sim','en'])
result = reader.readtext('image/nucleic-acid-20220419.jpg')
print(result)
项目目录结构：
4. 运行脚本
待检测的示例文件：
python easyocr_main.py
输出结果：
CUDA not available - defaulting to CPU. Note: This module is much faster with a GPU. [([[87, 89], [215, 89], [215, 149], [87, 149]], '姓名:', 0.9442024230957031), ([[888, 90], [990, 90], [990, 144], [888, 144]], '*辉', 0.3019644837089857), ([[89, 183], [309, 183], [309, 245], [89, 245]], '采样时间:', 0.9722564747406702), ([[522, 186], [992, 186], [992, 240], [522, 240]], '2022-04-2910:07.42', 0.5663659244856294), ([[89, 275], [309, 275], [309, 339], [89, 339]], '检测机构:', 0.928961890809416), ([[582, 272], [992, 272], [992, 336], [582, 336]], '亳州市华佗中医院', 0.9465525282106616), ([[89, 371], [305, 371], [305, 433], [89, 433]], '检测 时间:', 0.9039377327386754), ([[521, 373], [994, 373], [994, 427], [521, 427]], '2022-04-2916:31.06', 0.8148201099036173), ([[88, 465], [310, 465], [310, 525], [88, 525]], '检测项目:', 0.8194704755458363), ([[881, 463], [993, 463], [993, 523], [881, 523]], '核酸', 0.9890416189073347), ([[86, 558], [308, 558], [308, 618], [86, 618]], '检测结果:', 0.6191791236234889), ([[812, 550], [988, 550], [988, 622], [812, 622]], '[阴性]', 0.6905605792999268)]
命令行模式
easyocr.exe 一般被放置在 Python Scipts 目录下。
我的路径如是：D:ProgramPythonPython37Scriptseasyocr.exe
D:ProgramPythonPython37Scriptseasyocr -l ch_sim en -f image/nucleic-acid-20220419.jpg --detail=1 --gpu=False
输出结果：
([[87, 89], [215, 89], [215, 149], [87, 149]], '姓名:', 0.9442024230957031) ([[888, 90], [990, 90], [990, 144], [888, 144]], '*辉', 0.3019644837089857) ([[89, 183], [309, 183], [309, 245], [89, 245]], '采样时间:', 0.9722564747406702) ([[522, 186], [992, 186], [992, 240], [522, 240]], '2022-04-2910:07.42', 0.5663659244856294) ([[89, 275], [309, 275], [309, 339], [89, 339]], '检测机构:', 0.928961890809416) ([[582, 272], [992, 272], [992, 336], [582, 336]], '亳州市华佗中医院', 0.9465525282106616) ([[89, 371], [305, 371], [305, 433], [89, 433]], '检测时间:', 0.9039377327386754) ([[521, 373], [994, 373], [994, 427], [521, 427]], '2022-04-2916:31.06', 0.8148201099036173) ([[88, 465], [310, 465], [310, 525], [88, 525]], '检测项目:', 0.8194704755458363) ([[881, 463], [993, 463], [993, 523], [881, 523]], '核酸', 0.9890416189073347) ([[86, 558], [308, 558], [308, 618], [86, 618]], '检测结果:', 0.6191791236234889) ([[812, 550], [988, 550], [988, 622], [812, 622]], '[阴性]', 0.6905605792999268)
命令行 easyocr 使用帮助：
easyocr [-h] -l LANG [LANG ...] [--gpu {True,False}]
[--model_storage_directory MODEL_STORAGE_DIRECTORY]
[--user_network_directory USER_NETWORK_DIRECTORY]
[--recog_network RECOG_NETWORK]
[--download_enabled {True,False}] [--detector {True,False}]
[--recognizer {True,False}] [--verbose {True,False}]
[--quantize {True,False}] -f FILE
[--decoder {greedy,beamsearch,wordbeamsearch}]
[--beamWidth BEAMWIDTH] [--batch_size BATCH_SIZE]
[--workers WORKERS] [--allowlist ALLOWLIST]
[--blocklist BLOCKLIST] [--detail {0,1}]
[--rotation_info ROTATION_INFO] [--paragraph {True,False}]
[--min_size MIN_SIZE] [--contrast_ths CONTRAST_THS]
[--adjust_contrast ADJUST_CONTRAST]
[--text_threshold TEXT_THRESHOLD] [--low_text LOW_TEXT]
[--link_threshold LINK_THRESHOLD] [--canvas_size CANVAS_SIZE]
[--mag_ratio MAG_RATIO] [--slope_ths SLOPE_THS]
[--ycenter_ths YCENTER_THS] [--height_ths HEIGHT_THS]
[--width_ths WIDTH_THS] [--y_ths Y_THS] [--x_ths X_THS]
[--add_margin ADD_MARGIN]
Process EasyOCR.
optional arguments:
-h, --help show this help message and exit
-l LANG [LANG ...], --lang LANG [LANG ...]
for languages
--gpu {True,False} Using GPU (default: True)
--model_storage_directory MODEL_STORAGE_DIRECTORY
Directory for model (.pth) file
--user_network_directory USER_NETWORK_DIRECTORY
Directory for custom network files
--recog_network RECOG_NETWORK
Recognition networks
--download_enabled {True,False}
Enable Download
--detector {True,False}
Initialize text detector module
--recognizer {True,False}
Initialize text recognizer module
--verbose {True,False}
Print detail/warning
--quantize {True,False}
Use dynamic quantization
-f FILE, --file FILE input file
--decoder {greedy,beamsearch,wordbeamsearch}
decoder algorithm
--beamWidth BEAMWIDTH
size of beam search
--batch_size BATCH_SIZE
batch_size
--workers WORKERS number of processing cpu cores
--allowlist ALLOWLIST
Force EasyOCR to recognize only subset of characters
--blocklist BLOCKLIST
Block subset of character. This argument will be
ignored if allowlist is given.
--detail {0,1} simple output (default: 1)
--rotation_info ROTATION_INFO
Allow EasyOCR to rotate each text box and return the
one with the best confident score. Eligible values are
90, 180 and 270. For example, try [90, 180 ,270] for
all possible text orientations.
--paragraph {True,False}
Combine result into paragraph
--min_size MIN_SIZE Filter text box smaller than minimum value in pixel
--contrast_ths CONTRAST_THS
Text box with contrast lower than this value will be
passed into model 2 times. First is with original
image and second with contrast adjusted to
'adjust_contrast' value. The one with more confident
level will be returned as a result.
--adjust_contrast ADJUST_CONTRAST
target contrast level for low contrast text box
--text_threshold TEXT_THRESHOLD
Text confidence threshold
--low_text LOW_TEXT Text low-bound score
--link_threshold LINK_THRESHOLD
Link confidence threshold
--canvas_size CANVAS_SIZE
Maximum image size. Image bigger than this value will
be resized down.
--mag_ratio MAG_RATIO
Image magnification ratio
--slope_ths SLOPE_THS
Maximum slope (delta y/delta x) to considered merging.
Low value means tiled boxes will not be merged.
--ycenter_ths YCENTER_THS
Maximum shift in y direction. Boxes with different
level should not be merged.
--height_ths HEIGHT_THS
Maximum different in box height. Boxes with very
different text size should not be merged.
--width_ths WIDTH_THS
Maximum horizontal distance to merge boxes.
--y_ths Y_THS Maximum horizontal distance to merge boxes (when
paragraph = True).
--x_ths X_THS Maximum horizontal distance to merge boxes (when
paragraph = True).
--add_margin ADD_MARGIN
Extend bounding boxes in all direction by certain
value. This is important for language with complex
script (E.g. Thai).
参考文献
Github地址：GitHub - JaidedAI/EasyOCR: Ready-to-use OCR with 80+ supported languages and all popular writing scripts including Latin, Chinese, Arabic, Devanagari, Cyrillic and etc.https://github.com/JaidedAI/EasyOCR Python3行代码翻译70种语言！
