安装及使用 - RapidOCR 文档 #196
giscus[bot]
bot
Replies: 46 comments 54 replies
-
|
安装使用了，但得到TypeError: Object of type float32 is not JSON serializable错误。 |
Beta Was this translation helpful? Give feedback.
-
|
试用的rapidocr_web |
Beta Was this translation helpful? Give feedback.
-
|
麻烦给出具体运行环境和复现代码 |
Beta Was this translation helpful? Give feedback.
-
执行命令conda create -n rapidocr python=3.10
pip install rapidocr_web
rapidocr_web -ip 0.0.0.0 -p 9003控制台输出Package Version
-------------------- -----------
blinker 1.8.2
certifi 2024.7.4
charset-normalizer 3.3.2
click 8.1.7
coloredlogs 15.0.1
Flask 3.0.3
flatbuffers 24.3.25
humanfriendly 10.0
idna 3.7
itsdangerous 2.2.0
Jinja2 3.1.4
MarkupSafe 2.1.5
mpmath 1.3.0
numpy 1.26.4
onnxruntime 1.18.1
opencv-python 4.10.0.84
packaging 24.1
pillow 10.4.0
pip 24.0
protobuf 5.27.2
pyclipper 1.3.0.post5
PyYAML 6.0.1
rapidocr-onnxruntime 1.3.23
rapidocr-web 0.1.9
requests 2.32.3
setuptools 69.5.1
Shapely 1.8.5.post1
six 1.16.0
sympy 1.12.1
urllib3 2.2.2
Werkzeug 3.0.3
wheel 0.43.0python 3.10~12，都测试了，3.12安装rapidocr_web会报错，3.10和3.11运行图片检测时，都会报错: Successfully launched and visit https://0.0.0.0:9003 to view.
127.0.0.1 - - [05/Jul/2024 11:06:51] "GET / HTTP/1.1" 200 8176
127.0.0.1 - - [05/Jul/2024 11:06:51] "GET /static/css/main.css HTTP/1.1" 304 0
127.0.0.1 - - [05/Jul/2024 11:06:51] "GET /static/js/jquery-3.0.0.min.js HTTP/1.1" 304 0
[2024-07-05 11:07:15,867] ERROR in app: Exception on /ocr [POST]
Traceback (most recent call last):
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/flask/app.py", line 1473, in wsgi_app
response = self.full_dispatch_request()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/flask/app.py", line 882, in full_dispatch_request
rv = self.handle_user_exception(e)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/flask/app.py", line 880, in full_dispatch_request
rv = self.dispatch_request()
^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/flask/app.py", line 865, in dispatch_request
return self.ensure_sync(self.view_functions[rule.endpoint])(**view_args) # type: ignore[no-any-return]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/rapidocr_web/ocrweb.py", line 31, in ocr
ocr_res = processor(img_str)
^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/rapidocr_web/task.py", line 28, in __call__
return self.get_web_result(img, ocr_res, elapse)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/site-packages/rapidocr_web/task.py", line 51, in get_web_result
rec_res = json.dumps(rec_res, indent=2, ensure_ascii=False)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/__init__.py", line 238, in dumps
**kw).encode(obj)
^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 202, in encode
chunks = list(chunks)
^^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 430, in _iterencode
yield from _iterencode_list(o, _current_indent_level)
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 326, in _iterencode_list
yield from chunks
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 326, in _iterencode_list
yield from chunks
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 439, in _iterencode
o = _default(o)
^^^^^^^^^^^
File "/home/longdexin/miniconda3/envs/test/lib/python3.11/json/encoder.py", line 180, in default
raise TypeError(f'Object of type {o.__class__.__name__} '
TypeError: Object of type float32 is not JSON serializable |
Beta Was this translation helpful? Give feedback.
-
|
收到，周末看一下 |
Beta Was this translation helpful? Give feedback.
-
|
已经在 |
Beta Was this translation helpful? Give feedback.
-
ImportError: DLL load failed while importing onnxruntime_pybind11_state: 动态链接库(DLL)初始化例程失败。 使用onnxruntime-gpu时出错 |
Beta Was this translation helpful? Give feedback.
-
|
首先不建议使用rapidocr_onnxruntime gpu版哈，这个不能有加速效果。建议用 |
Beta Was this translation helpful? Give feedback.
-
|
你好，我实际测试下来，发现CPU上rapidocr_onnxruntime性能比rapidocr_paddle好，但是这个差异主要是默认的cpu_math_library_num_threads没有设置导致的。具体原因是：以下这个分支里的设置线程数没有跑进去，导致paddle CPU推理的默认性能很差。 |
Beta Was this translation helpful? Give feedback.
-
|
收到，不知是否愿意提个PR来修复这个问题呢? |
Beta Was this translation helpful? Give feedback.
-
|
同一张图片use_det参数设置为True时可以正常识别文字，但是设置为False时就会识别到内容为空，这是为什么 |
Beta Was this translation helpful? Give feedback.
-
|
首先要明白这个参数是干嘛的？ |
Beta Was this translation helpful? Give feedback.
-
|
您好。我怎么来提高这个文本识别的经度。怎么样微调下？多谢 |
Beta Was this translation helpful? Give feedback.
-
|
RapidOCR不支持微调，要想提高文本识别精度，一是可以换已有高精度模型(docs)，二是基于PaddleOCR来微调模型。 |
Beta Was this translation helpful? Give feedback.
-
|
想要支持GPU推理的rapidocr_api |
Beta Was this translation helpful? Give feedback.
-
|
这个加到TODO吧，后期放开绑定的rapiddocr库，着急的话，可以直接修改源码。 |
Beta Was this translation helpful? Give feedback.
-
|
现在rapiddocr_api支持gpu推理了吗 |
Beta Was this translation helpful? Give feedback.
-
|
CnOCR就支持GPU的API，不过就是识别率不满意，https://cnocr.readthedocs.io/zh-cn/stable/install/ |
Beta Was this translation helpful? Give feedback.
-
|
请问下大佬，集成到我们Springboot项目中后，在本地（windows环境）调试ocr识别没问题。但是到了测试环境（linux环境）调ocr识别会导致容器立马崩溃重启（退出码139）是什么问题呢 |
Beta Was this translation helpful? Give feedback.
-
|
还有这个问题： |
Beta Was this translation helpful? Give feedback.
-
|
用的哪个编程语言的版本呀？请详细说明环境，代码啥的 |
Beta Was this translation helpful? Give feedback.
-
|
请问解决了吗，我也碰到了这个问题 |
Beta Was this translation helpful? Give feedback.
-
|
我也遇到这个问题了，不知道咋个解决了，java代码InferenceEngine engine = InferenceEngine.getInstance(Model.ONNX_PPOCR_V3); |
Beta Was this translation helpful? Give feedback.
-
|
rapidocr_web 支持通过API调用吗？像rapidocr_api一样。 |
Beta Was this translation helpful? Give feedback.
-
|
为啥不直接用rapidocr_api呢？ |
Beta Was this translation helpful? Give feedback.
-
|
你发了，用 rapidocr_api 也可以啊，不过我想要在本地桌面上使用，就是说我要像 rapidocr_web 那样打包，但 rapidocr_api 要怎么打包？ |
Beta Was this translation helpful? Give feedback.
-
|
能给写个 rapidocr_api 打包的教程吗？ |
Beta Was this translation helpful? Give feedback.
-
|
您好博主, 我目前检测模型使用PPOCRv3训练的, 然后转化为onnx后,发现 未转化onnx的检测推理模型, 可以检测到图片, 转化到onnx后 检测不到文字了, 识别模型是基于PPOCRv4做的训练, 转化后精度没损失, 可以识别, |
Beta Was this translation helpful? Give feedback.
-
|
安装时默认会安装哪个引擎？ |
Beta Was this translation helpful? Give feedback.
-
|
安装名字就已经决定了是哪个引擎。 pip install rapidocr_onnxruntime
pip install rapidocr_openvino |
Beta Was this translation helpful? Give feedback.
-
|
rapidocr_paddle包是基于PaddlePaddle框架作为推理引擎的，支持CPU和GPU上推理，推荐GPU端使用。原因是ONNXRuntime和OpenVINO在GPU上支持较差。 |
Beta Was this translation helpful? Give feedback.
-
|
这话后面是不是有点矛盾。支持GPU，原因在GPU 支持性较差 ？？ |
Beta Was this translation helpful? Give feedback.
-
|
不矛盾。onnxruntime本身是支持GPU推理的，有onnxruntime-gpu版本的。但是对于OCR场景，其输入图像shape是动态的，onnxruntime对这个支持较差，因此GPU端提速十分有限，甚至比CPU慢。 |
Beta Was this translation helpful? Give feedback.
-
|
rapidocr_paddle会不会像paddle一样，有显存泄露问题？本人在很多版本paddle+paddleocr上做了测试工作，都有显存泄露问题。 |
Beta Was this translation helpful? Give feedback.
-
|
会的，等v2.0.0的torch版本吧 |
Beta Was this translation helpful? Give feedback.
-
|
大哥，大概什么时候出来？很急 |
Beta Was this translation helpful? Give feedback.
-
|
去看一下PR，已经有了，只是没有合并
…---- 回复的原邮件 ----
| 发件人 | ***@***.***> |
| 发送日期 | 2025年01月22日 17:41 |
| 收件人 | RapidAI/RapidOCR ***@***.***> |
| 抄送人 | SWHL ***@***.***>,
State ***@***.***> |
| 主题 | Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196) |
大哥，大概什么时候出来？很急
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you modified the open/close state.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
为什么将以 docker 部署的 rapidocr_api 中的参数文件修改成以下配置就无法识别出文字了呢？只要将 Global:
use_det: false
use_cls: false
use_rec: true |
Beta Was this translation helpful? Give feedback.
-
|
现阶段的OCR流程是先用det将文本位置检测出来，后续才能进行识别 |
Beta Was this translation helpful? Give feedback.
-
|
好的，感谢解答 |
Beta Was this translation helpful? Give feedback.
-
|
所以现在一定要使用use det = True 是吗 |
Beta Was this translation helpful? Give feedback.
-
|
对于多行文本识别，是一定要用的。单行文本可以不用 |
Beta Was this translation helpful? Give feedback.
-
|
请问用paddlepaddle推理引擎，支持在npu上推理吗 |
Beta Was this translation helpful? Give feedback.
-
|
这个paddle框架是支持的，但是在rapidocr中暂时没有适配。欢迎贡献代码哈
…---- 回复的原邮件 ----
| 发件人 | ***@***.***> |
| 发送日期 | 2025年07月02日 17:14 |
| 收件人 | RapidAI/RapidOCR ***@***.***> |
| 抄送人 | SWHL ***@***.***>,
State ***@***.***> |
| 主题 | Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196) |
请问用paddlepaddle推理引擎，支持在npu上推理吗
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you modified the open/close state.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
你好！rapidocr_api可以调用gpu吗？为啥我这里gpu的调用为0 |
Beta Was this translation helpful? Give feedback.
-
|
nvidia-smi是有输出的，cuda也安装了，会有什么原因导致用不了gpu呢？ |
Beta Was this translation helpful? Give feedback.
-
|
torch怎么用rapidocr_api呢？ |
Beta Was this translation helpful? Give feedback.
-
|
请问如何在npu上如300i duo上运行呢？ |
Beta Was this translation helpful? Give feedback.
-
|
参见文档：docs |
Beta Was this translation helpful? Give feedback.
-
|
初始化RapidOCR时，Det.lang_type设置为LangDet.MULTI, 会报错 File "/app/paddle_server/core/ocr/rapid.py", line 18, in |
Beta Was this translation helpful? Give feedback.
-
|
烦请给出完整的初始化代码。我猜测是因为不支持这个模型。这个docs中写着不支持的，运行的话，就会报上面错误哈。 |
Beta Was this translation helpful? Give feedback.
-
问题描述 / Problem Description传入ch_en_num.jpg这张图，输出结果是None 运行环境 / Runtime Environmentcuda11.7 复现代码 / Reproduction Codeimport cv2
from rapidocr_paddle import RapidOCR
# 注意这里的参数
engine = RapidOCR(det_use_cuda=True, cls_use_cuda=True, rec_use_cuda=True)
image_path = "xxx/ch_en_num.jpg"
result, elapse_list = engine(image_path)
print(result) |
Beta Was this translation helpful? Give feedback.
-
|
试试paddlepaddle cpu版呢 或者升级paddlepaddle -GPU最新版 |
Beta Was this translation helpful? Give feedback.
-
|
支持参数设置吗 比如模型 返回单字符等 |
Beta Was this translation helpful? Give feedback.
-
|
支持的。多看看文档，都有写的。 |
Beta Was this translation helpful? Give feedback.
-
|
模型环境变量似乎无效呢？即使指定了还是默认的： |
Beta Was this translation helpful? Give feedback.
-
|
看了下rapidapi的代码，需要检测到3个变量才会使用自定义模型，所以Windows下正确的使用方法是：set det_model_path=D:\RapidOCR\ch_PP-OCRv5_server_det.onnx |
Beta Was this translation helpful? Give feedback.
-
|
rapidocr 支持 onnxruntime-cann 下面运行吗，我看有cann相关的配置。我在onnxruntime-cann 1.23.0 下面测试报错 |
Beta Was this translation helpful? Give feedback.
-
|
支持的。但是需要将除 |
Beta Was this translation helpful? Give feedback.
-
|
您好，根据上面进行了rapidocr+pytorch测试，发现同一张图片用v4推理结果正确，用v5推理结果完全对不上，这是否是后处理导致的？用的是“绿洲仕格维花园公寓”这张图片。试过很多种方式推理pp-ocrv5，发现使用v4的推理去进行v5推理，都会对结果造成影响，是不是v4的推理不能完整契合v5？ |
Beta Was this translation helpful? Give feedback.
-
|
没看太懂。请贴一下你的代码 |
Beta Was this translation helpful? Give feedback.
-
|
from rapidocr import EngineType, RapidOCR, OCRVersion engine = RapidOCR( img_url = "general_ocr_rec_001.png" from rapidocr import EngineType, RapidOCR, OCRVersion engine = RapidOCR( img_url = "general_ocr_rec_001.png" 这两个使用的都是同一张图片，结果完全不同 |
Beta Was this translation helpful? Give feedback.
-
|
from rapidocr import EngineType, RapidOCR, OCRVersion engine = RapidOCR( img_url = "general_ocr_rec_001.png" |
Beta Was this translation helpful? Give feedback.
-
|
一般识别的图像只有一行文本这种，就不要用det了。这种det容易漏检
…---- 回复的原邮件 ----
| 发件人 | ***@***.***> |
| 发送日期 | 2025年09月08日 17:56 |
| 收件人 | RapidAI/RapidOCR ***@***.***> |
| 抄送人 | SWHL ***@***.***>,
State ***@***.***> |
| 主题 | Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196) |
from rapidocr import EngineType, RapidOCR, OCRVersion
engine = RapidOCR(
params={
"Det.engine_type": EngineType.TORCH,
"Det.ocr_version": OCRVersion.PPOCRV5,
"Cls.engine_type": EngineType.TORCH,
"Cls.ocr_version": OCRVersion.PPOCRV5,
"Rec.engine_type": EngineType.TORCH,
"Rec.ocr_version": OCRVersion.PPOCRV5,
"EngineConfig.torch.use_cuda": True, # 使用torch GPU版推理
"EngineConfig.torch.gpu_id": 0, # 指定GPU id
"EngineConfig.torch.torch_dtype": "float16",
}
)
img_url = "general_ocr_rec_001.png"
result = engine(img_url, use_det=False, use_cls=False, use_rec=True)
print(result.txts)
经测试，关闭det模型，输出结果正确，应该是det没有正确识别到文本，文本框不对导致的
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you modified the open/close state.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
好的谢谢，问下rapidOCR支持jetson nano板卡部署ocrv5推理加速吗？毕竟板卡性能低，但是实际产线速度不满，一秒差不多5-10张才行。
…---原始邮件---
发件人: ***@***.***>
发送时间: 2025年9月8日(周一) 晚上7:41
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196)
一般识别的图像只有一行文本这种，就不要用det了。这种det容易漏检
---- 回复的原邮件 ----
| 发件人 | ***@***.***> |
| 发送日期 | 2025年09月08日 17:56 |
| 收件人 | RapidAI/RapidOCR ***@***.***> |
| 抄送人 | SWHL ***@***.***>,
State ***@***.***> |
| 主题 | Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196) |
from rapidocr import EngineType, RapidOCR, OCRVersion
engine = RapidOCR(
params={
"Det.engine_type": EngineType.TORCH,
"Det.ocr_version": OCRVersion.PPOCRV5,
"Cls.engine_type": EngineType.TORCH,
"Cls.ocr_version": OCRVersion.PPOCRV5,
"Rec.engine_type": EngineType.TORCH,
"Rec.ocr_version": OCRVersion.PPOCRV5,
"EngineConfig.torch.use_cuda": True, # 使用torch GPU版推理
"EngineConfig.torch.gpu_id": 0, # 指定GPU id
"EngineConfig.torch.torch_dtype": "float16",
}
)
img_url = "general_ocr_rec_001.png"
result = engine(img_url, use_det=False, use_cls=False, use_rec=True)
print(result.txts)
经测试，关闭det模型，输出结果正确，应该是det没有正确识别到文本，文本框不对导致的
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you modified the open/close state.Message ID: ***@***.***>
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
支持的。刚查了一下jetson nano，PyTorch和PaddlePaddle都能在上面使用。需要正确安装GPU版的，之后，就能推理加速了。 |
Beta Was this translation helpful? Give feedback.
-
|
非常感谢，他这个是paddle2，想用ocrv5必须要3以上才行，这就要自己编译，编译的话太麻烦，而且不知道会出现什么问题，不行就只能转trt引擎用pycuda推理了
…---原始邮件---
发件人: ***@***.***>
发送时间: 2025年9月8日(周一) 晚上9:20
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196)
支持的。刚查了一下jetson nano，PyTorch和PaddlePaddle都能在上面使用。需要正确安装GPU版的，之后，就能推理加速了。
参考博客类似博客：link
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
大部分情况，v4就能满足了。可以先试试v4 |
Beta Was this translation helpful? Give feedback.
-
|
现在就是用的fastdeploy推理v4，但是会出现漏识别的情况，用v5测试就不会出现这种情况，但是v5要fastdeploy2.0才行，官方文档现在也没有端侧的信息了，全是大模型，不知道还支持端侧不，编译还难，各种问题
…---原始邮件---
发件人: ***@***.***>
发送时间: 2025年9月8日(周一) 晚上9:39
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [RapidAI/RapidOCR] 安装及使用 - RapidOCR 文档 (Discussion #196)
大部分情况，v4就能满足了。可以先试试v4
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
and others
Uh oh!
There was an error while loading. Please reload this page.
-
安装及使用 - RapidOCR 文档
Awesome OCR multiple programing languages toolkits based on ONNXRuntime, OpenVION and PaddlePaddle.
https://rapidai.github.io/RapidOCRDocs/install_usage/rapidocr_web/rapidocr_web/
Beta Was this translation helpful? Give feedback.
All reactions