PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/doc_preprocessor

[giscus[bot]]

PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/doc_preprocessor

https://paddlepaddle.github.io/PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/doc_preprocessor.html

[husthzy]

API中的接口名称写错了，使用官方docker部署，接口名称应该是：/document-preprocessing

[Bobholamovic]

感谢指出，我们将尽快修复这个问题

[Bobholamovic]

该问题已在develop分支以及最新的release分支修复～

[GY13648114184]

怎么解决我在用paddlenlp中的uie抽取时对单元格中文本有换行的情况只能识别到第一行的问题呢？

[cuicheng01]

这种情况，可以试试chatocrv3？

[mynameisQQQ]

严重性 代码 说明 项目 文件 行 禁止显示状态 详细信息
错误 C2039 "DisableMKLDNN": 不是 "paddle::AnalysisConfig" 的成员 D:\Optical Character Recognition\PaddleOCR-main\deploy\cpp_infer\out\build\release\cpp_infer D:\Optical Character Recognition\PaddleOCR-main\deploy\cpp_infer\src\ocr_cls.cpp 135
这个错误怎么解决

[lxridewind]

求教各位大神，这个产线里5类模型，经过高性能推理安装调试之后，就剩下 文本图像矫正模块UVDoc 最耗时，前一步 文档图像方向分类模块 已经知道输入图片是0/90/180/270角度了，能不能插入自定义开发替换掉 UVDoc 过程，因为我只需要一个简单的旋转图像就够了。
如果直接禁用UVDoc没有旋转矫正，那么最后识别返回的文字快顺序有问题不连贯，影响喂给LLM大模型分析抽取。 如果说直接用 文档场景信息抽取V3产线，则响应的较慢不够实时快捷。

[Hisir0909]

最简单的实现应该是 前置文档图像预处理产线(关闭UVDoc), 然后得到结果自行旋转图像, 旋转后的数据发给 通用OCR产线(关闭预处理子产线) 不知道符合你的需求不.

[junxiaoge]

识别结果可以返回单个字母的坐标吗，比如“928”文字，返回三个坐标

[Ceezy]

目前原始图片、翻转后的图片、扭曲矫正后的图片是一起输出的，是否能分开多个图片输出？

[FengTingLiuUmi]

从res里分别取output_img等几个数据就行。

[Tomhardy13]

下了文本图像矫正模块的UVDoc的推理模型，只有inference.json没inference.pdmodel，无法paddlelite转换，下训练模型又找不到UVDoc的参数文件，所以目前UVDoc是无法转成nb模型吗？

[hssxin]

root@localhost:/work# paddlex --pipeline OCR \

    --input https://wx4.sinaimg.cn/mw690/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg \
    --use_doc_orientation_classify False \
    --use_doc_unwarping False \
    --use_textline_orientation False \
    --save_path ./output \
    --device npu:0

Creating model: ('PP-LCNet_x1_0_doc_ori', None)
Using official model (PP-LCNet_x1_0_doc_ori), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.09it/s]
I0827 13:07:46.373286 37003 init.cc:238] ENV [CUSTOM_DEVICE_ROOT]=/usr/local/lib/python3.10/dist-packages/paddle_custom_device
I0827 13:07:46.373381 37003 init.cc:146] Try loading custom device libs from: [/usr/local/lib/python3.10/dist-packages/paddle_custom_device]
I0827 13:07:47.009894 37003 custom_device_load.cc:51] Succeed in loading custom runtime in lib: /usr/local/lib/python3.10/dist-packages/paddle_custom_device/libpaddle-custom-npu.so
I0827 13:07:47.009954 37003 custom_device_load.cc:58] Skipped lib [/usr/local/lib/python3.10/dist-packages/paddle_custom_device/libpaddle-custom-npu.so]: no custom engine Plugin symbol in this lib.
I0827 13:07:47.014819 37003 custom_kernel.cc:68] Succeed in loading 359 custom kernel(s) from loaded lib(s), will be used like native ones.
I0827 13:07:47.015035 37003 init.cc:158] Finished in LoadCustomDevice with libs_path: [/usr/local/lib/python3.10/dist-packages/paddle_custom_device]
I0827 13:07:47.015079 37003 init.cc:244] CustomDevice: npu, visible devices count: 6
Creating model: ('UVDoc', None)
Using official model (UVDoc), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.62it/s]
Creating model: ('PP-LCNet_x1_0_textline_ori', None)
Using official model (PP-LCNet_x1_0_textline_ori), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.94it/s]
Creating model: ('PP-OCRv5_server_det', None)
Using official model (PP-OCRv5_server_det), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.71it/s]
Creating model: ('PP-OCRv5_server_rec', None)
Using official model (PP-OCRv5_server_rec), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:01<00:00, 2.75it/s]
Connecting to https://wx4.sinaimg.cn/mw690/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg ...
Downloading 008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg ...
[==================================================] 100.00%
{'res': {'input_path': '/root/.paddlex/predict_input/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg', 'page_index': None, 'model_settings': {'use_doc_preprocessor': False, 'use_textline_orientation': False}, 'dt_polys': array([], dtype=float64), 'text_det_params': {'limit_side_len': 64, 'limit_type': 'min', 'thresh': 0.3, 'max_side_limit': 4000, 'box_thresh': 0.6, 'unclip_ratio': 1.5}, 'text_type': 'general', 'text_rec_score_thresh': 0.0, 'return_word_box': False, 'rec_texts': [], 'rec_scores': array([], dtype=float64), 'rec_polys': array([], dtype=float64), 'rec_boxes': array([], dtype=float64)}}
free(): invalid pointer

---

C++ Traceback (most recent call last):

No stack trace in paddle, may be caused by external reasons.

---

Error Message Summary:

FatalError: Process abort signal is detected by the operating system.
[TimeInfo: *** Aborted at 1756271311 (unix time) try "date -d @1756271311" if you are using GNU date ***]
[SignalInfo: *** SIGABRT (@0x908b) received by PID 37003 (TID 0xffffb909c010) from PID 37003 ***]

Aborted (core dumped)
root@localhost:/work# paddlex --pipeline OCR --input general_ocr_002.png --use_doc_orientation_classify False --use_doc_unwarping False --use_textline_orientation False --save_path ./output
Creating model: ('PP-LCNet_x1_0_doc_ori', None)
Using official model (PP-LCNet_x1_0_doc_ori), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 0%| | 0.00/5.00 [00:00<?, ?it/s]^Z
[23]+ Stopped paddlex --pipeline OCR --input general_ocr_002.png --use_doc_orientation_classify False --use_doc_unwarping False --use_textline_orientation False --save_path ./output
root@localhost:/work# paddlex --pipeline OCR \

    --input https://wx4.sinaimg.cn/mw690/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg \
    --use_doc_orientation_classify False \
    --use_doc_unwarping False \
    --use_textline_orientation False \
    --save_path ./output

Creating model: ('PP-LCNet_x1_0_doc_ori', None)
Using official model (PP-LCNet_x1_0_doc_ori), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.97it/s]
I0827 13:09:19.245047 41819 init.cc:238] ENV [CUSTOM_DEVICE_ROOT]=/usr/local/lib/python3.10/dist-packages/paddle_custom_device
I0827 13:09:19.245098 41819 init.cc:146] Try loading custom device libs from: [/usr/local/lib/python3.10/dist-packages/paddle_custom_device]
I0827 13:09:19.921327 41819 custom_device_load.cc:51] Succeed in loading custom runtime in lib: /usr/local/lib/python3.10/dist-packages/paddle_custom_device/libpaddle-custom-npu.so
I0827 13:09:19.921386 41819 custom_device_load.cc:58] Skipped lib [/usr/local/lib/python3.10/dist-packages/paddle_custom_device/libpaddle-custom-npu.so]: no custom engine Plugin symbol in this lib.
I0827 13:09:19.926184 41819 custom_kernel.cc:68] Succeed in loading 359 custom kernel(s) from loaded lib(s), will be used like native ones.
I0827 13:09:19.926398 41819 init.cc:158] Finished in LoadCustomDevice with libs_path: [/usr/local/lib/python3.10/dist-packages/paddle_custom_device]
I0827 13:09:19.926447 41819 init.cc:244] CustomDevice: npu, visible devices count: 6
Creating model: ('UVDoc', None)
Using official model (UVDoc), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 7.50it/s]
Creating model: ('PP-LCNet_x1_0_textline_ori', None)
Using official model (PP-LCNet_x1_0_textline_ori), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 7.20it/s]
Creating model: ('PP-OCRv5_server_det', None)
Using official model (PP-OCRv5_server_det), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:00<00:00, 6.66it/s]
Creating model: ('PP-OCRv5_server_rec', None)
Using official model (PP-OCRv5_server_rec), the model files will be automatically downloaded and saved in /root/.paddlex/official_models.
Processing 5 items: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 5.00/5.00 [00:01<00:00, 3.94it/s]
Connecting to https://wx4.sinaimg.cn/mw690/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg ...
Downloading 008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg ...
[==================================================] 100.00%
{'res': {'input_path': '/root/.paddlex/predict_input/008CmN0fly1hv6jt7i2nsj30u011r7wh.jpg', 'page_index': None, 'model_settings': {'use_doc_preprocessor': False, 'use_textline_orientation': False}, 'dt_polys': array([[[ 62, 0],
...,
[ 62, 27]],

   ...,

   [[283, 841],
    ...,
    [283, 862]]], dtype=int16), 'text_det_params': {'limit_side_len': 64, 'limit_type': 'min', 'thresh': 0.3, 'max_side_limit': 4000, 'box_thresh': 0.6, 'unclip_ratio': 1.5}, 'text_type': 'general', 'textline_orientation_angles': array([-1, ..., -1]), 'text_rec_score_thresh': 0.0, 'return_word_box': False, 'rec_texts': ['上海虹桥站', 'G28', '北京南站', 'Shanghaihongqiao', 'Beijingnan', '2024年10月25日19:00开', '14车12F号', '￥626.0元', '惠', '二等座', '仅供报销使用', 'Z84S058947', '检票：10A、10B', '北京南站', 'G11', '上海虹桥站', 'Beijingnan', 'Shanghaihongqiao', '2024年10月06日12:00开', '06车04F号', '￥662.0元', '二等座', '仅供报销使用', 'Z84S058946', '检票：1AB', '上海虹桥站', 'G22', '北京南站', 'Shanghaihongqiao', 'Beijingnan', '2024年10月01日16:00开', '06车12F号', '￥662.0元', '二等座', '检票：1AB', 'Z88G008231', 'G28', '北京南站', '上海虹桥站', 'Shanghaihongqiao', 'Beijingnan', '2024年09月14日19:00开', '07车08F号', '惠', '二等座', '￥626.0元', '仅供报销使用', '@午夜不睡说个机'], 'rec_scores': array([0.9968766 , ..., 0.99799263]), 'rec_polys': array([[[ 62,   0],
    ...,
    [ 62,  27]],

   ...,

   [[283, 841],
    ...,
    [283, 862]]], dtype=int16), 'rec_boxes': array([[ 62, ...,  27],
   ...,
   [283, ..., 862]], dtype=int16)}}

corrupted size vs. prev_size

[hssxin]

无法使用NPU进行OCR识别？

[pcupso]

高性能部署OCR运行成功后，调用8000端口接口测试，报错如下：
[ERROR] ultra_infer/runtime/backends/tensorrt/trt_backend.cc(445)::SetInputs TRTBackend SetInputs not find name:x

---

C++ Traceback (most recent call last):

No stack trace in paddle, may be caused by external reasons.

---

Error Message Summary:

FatalError: Process abort signal is detected by the operating system.
[TimeInfo: *** Aborted at 1756716977 (unix time) try "date -d @1756716977" if you are using GNU date ***]
[SignalInfo: *** SIGABRT (@0xa7b) received by PID 2683 (TID 0x72e393c32000) from PID 2683 ***]

E0901 08:56:18.371417 7 python.cc:1944] Stub process is unhealthy and it will be restarted.

版本是docker gpu3.2、3.1，都运行不起来，帮忙看看什么原因，在线等，着急

[pcupso]

占用正常 1349MiB / 8188MiB ，详细如下：

Fri Sep 5 06:13:15 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.181 Driver Version: 570.181 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 4060 Off | 00000000:01:00.0 On | N/A |
| 0% 49C P8 N/A / 115W | 1349MiB / 8188MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 43485 C tritonserver 156MiB |
| 0 N/A N/A 44248 C ...on/triton_python_backend_stub 1170MiB |
+-----------------------------------------------------------------------------------------+

[Bobholamovic]

我在自己的机器上用3.2版本的最新镜像和SDK无法复现这个问题……请问不开高性能推理的话，可以正常运作吗？

[pcupso]

不开高性能可以

[pcupso]

我看问题报错也在tensorRT这边：

[ERROR] ultra_infer/runtime/backends/tensorrt/trt_backend.cc(636)::BuildTrtEngine Failed to call buildSerializedNetwork().

[Bobholamovic]

建议可以关注一下引擎构建期间的显存占用，引擎构建有可能消耗比较多的显存，不排除会超过8GB的上限