Install PaddleOCR on AGX (Python 3.8)

We need to install several Python packages.

Paddle
PaddleOCR
layoutparser

ENV

AGX device
Ubuntu OS 18.04
Jetpack v4.6
Python3.8
GCC v7.5

Steps

python3.8 -m pip install setuptools

apt-get update
apt-get install -y libgeos-dev
apt-get install -y python3-sklearn
apt-get install -y libxml2-dev libxslt-dev python3-dev
apt-get install -y python3-lxml
apt-get install -y patchelf

python3.8 -m pip install scikit-build cmake

### try to install the scikit-image package
apt-get install -y libaec-dev libblosc-dev libffi-dev libbrotli-dev libboost-all-dev libbz2-dev
apt-get install -y libgif-dev libopenjp2-7-dev liblcms2-dev libjpeg-dev libjxr-dev liblz4-dev liblzma-dev libpng-dev libsnappy-dev libwebp-dev libzopfli-dev libzstd-dev
apt-get install -y python3.8-dev

### from 1.13 - 1.21, opencv-python 4.5.4.60 has requirement numpy>=1.19.3
python3.8 -m pip install pip --upgrade
python3.8 -m pip install numpy==1.19.3

### PyWavelets-1.2.0 imageio-2.13.5 networkx-2.6.3 pillow-8.4.0 scikit-image-0.19.1 scipy-1.7.3 tifffile-2021.11.2
python3.8 -m pip install scikit-image==0.19.1

python3.8 -m pip install shellcheck-py==0.8.0.1
python3.8 -m pip install -U https://paddleocr.bj.bcebos.com/whl/layoutparser-0.0.0-py3-none-any.whl

python3.8 -m pip install paddleocr==2.3.0.2

### I will disable this line 
# python3.8 -m pip install "paddleocr>=2.0.1"

Depend, if you already installed.

python3.8 -m pip install opencv-python

Install Paddle

Reference: https://www.codeleading.com/article/34695546218/

Install NCCL tool.

git clone https://github.com/NVIDIA/nccl.git
cd nccl
make -j6
sudo make install

Start to build Paddle packages.

git clone https://github.com/PaddlePaddle/Paddle.git
cd Paddle
git checkout release/2.0

Create a script named as nx_cmake.sh I was using python3.8 to build it.

#nx_cmake.sh
if [ ! -d "build" ]; then
  mkdir build
fi

cd build

cmake .. \
  -DWITH_CONTRIB=OFF \
  -DWITH_MKL=OFF  \
  -DWITH_MKLDNN=OFF \
  -DWITH_AVX=OFF \
  -DWITH_GPU=ON \
  -DWITH_TESTING=OFF \
  -DCMAKE_BUILD_TYPE=Release \
  -DON_INFER=ON \
  -DWITH_PYTHON=ON  \
  -DPY_VERSION=3.8  \
  -DWITH_XBYAK=OFF  \
  -DWITH_NV_JETSON=ON

(Can add one more option for AGX -DCMAKE_CUDA_ARCHITECTURES=72)

Let's start it

sh ./nx_cmake.sh

cd build
make -j6

After complete the building process, let's install the whl package by pip.

cd /paddle/build/python/dist
pip3 install -U（whl包的名字）

Done~

Troubleshooting:

During building the package from source, we encountered this issue:

Error output

copying /root/Paddle/paddle/fluid/platform/dynload/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/platform/dynload
copying /root/Paddle/paddle/fluid/platform/details/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/platform/details
copying /root/Paddle/build/paddle/fluid/platform/*.pb.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/platform
copying /root/Paddle/paddle/fluid/string/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/string
copying /root/Paddle/paddle/fluid/string/tinyformat/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/string/tinyformat
copying /root/Paddle/paddle/fluid/imperative/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/imperative
copying /root/Paddle/paddle/fluid/imperative/jit/*.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/imperative/jit
copying /root/Paddle/build/paddle/fluid/pybind/pybind.h -> /root/Paddle/build/paddle_install_dir/paddle/fluid/pybind
copying /root/Paddle/build/paddle_inference_install_dir/third_party -> /root/Paddle/build/paddle_install_dir
copying /root/Paddle/build/CMakeCache.txt -> /root/Paddle/build/paddle_install_dir
[ 99%] Built target fluid_lib_dist
[ 99%] Building CXX object paddle/fluid/pybind/CMakeFiles/paddle_pybind.dir/nccl_wrapper_py.cc.o
[ 99%] Linking CXX shared library libpaddle_pybind.so
[ 99%] Built target paddle_pybind
[100%] Generating paddle/fluid/core_noavx.so
[100%] Built target copy_paddle_pybind
[100%] Packing whl packages------>>>
fatal: no tag exactly matches '462ee101224a7d2ac2ea1a88d41ef90e341a98a4'
sh: 1: patchelf: not found
Traceback (most recent call last):
  File "setup.py", line 489, in <module>
    raise Exception("patch core_noavx.%s failed, command: %s" % (ext_name, command))
Exception: patch core_noavx..so failed, command: patchelf --set-rpath '$ORIGIN/../libs/' /root/Paddle/build/python/paddle/fluid/core_noavx.so
python/CMakeFiles/paddle_python.dir/build.make:2113: recipe for target 'python/build/.timestamp' failed
make[2]: *** [python/build/.timestamp] Error 1
CMakeFiles/Makefile2:104564: recipe for target 'python/CMakeFiles/paddle_python.dir/all' failed
make[1]: *** [python/CMakeFiles/paddle_python.dir/all] Error 2
Makefile:135: recipe for target 'all' failed
make: *** [all] Error 2

Solution:

apt-get install patchelf

AttributeError: module layoutparser has no attribute PaddleDetectionLayoutModel

Error message:

root@fcf09b4b8614:~/Paddle/build/python/dist# python3.8
Python 3.8.12 (default, Sep 10 2021, 00:16:05) 
[GCC 7.5.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from paddleocr import PPStructure,draw_structure_result,save_structure_res
>>> table_engine = PPStructure(show_log=True)
Namespace(benchmark=False, cls_batch_num=6, cls_image_shape='3, 48, 192', cls_model_dir=None, cls_thresh=0.9, cpu_threads=10, det=True, det_algorithm='DB', det_db_box_thresh=0.6, det_db_score_mode='fast', det_db_thresh=0.3, det_db_unclip_ratio=1.5, det_east_cover_thresh=0.1, det_east_nms_thresh=0.2, det_east_score_thresh=0.8, det_limit_side_len=960, det_limit_type='max', det_model_dir='/root/.paddleocr/2.3.0.2/ocr/det/ch/ch_PP-OCRv2_det_infer', det_pse_box_thresh=0.85, det_pse_box_type='box', det_pse_min_area=16, det_pse_scale=1, det_pse_thresh=0, det_sast_nms_thresh=0.2, det_sast_polygon=False, det_sast_score_thresh=0.5, drop_score=0.5, e2e_algorithm='PGNet', e2e_char_dict_path='./ppocr/utils/ic15_dict.txt', e2e_limit_side_len=768, e2e_limit_type='max', e2e_model_dir=None, e2e_pgnet_mode='fast', e2e_pgnet_polygon=True, e2e_pgnet_score_thresh=0.5, e2e_pgnet_valid_set='totaltext', enable_mkldnn=False, gpu_mem=500, help='==SUPPRESS==', image_dir=None, ir_optim=True, label_list=['0', '180'], lang='ch', layout_path_model='lp://PubLayNet/ppyolov2_r50vd_dcn_365e_publaynet/config', max_batch_size=10, max_text_length=25, min_subgraph_size=15, ocr_version='PP-OCRv2', output='./output/table', precision='fp32', process_id=0, rec=True, rec_algorithm='CRNN', rec_batch_num=6, rec_char_dict_path='/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/utils/ppocr_keys_v1.txt', rec_image_shape='3, 32, 320', rec_model_dir='/root/.paddleocr/2.3.0.2/ocr/rec/ch/ch_PP-OCRv2_rec_infer', save_log_path='./log_output/', show_log=True, structure_version='STRUCTURE', table_char_dict_path='/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/utils/dict/table_structure_dict.txt', table_char_type='en', table_max_len=488, table_model_dir='/root/.paddleocr/2.3.0.2/ocr/table/en_ppocr_mobile_v2.0_table_structure_infer', total_process_num=1, type='ocr', use_angle_cls=False, use_dilation=False, use_gpu=True, use_mp=False, use_onnx=False, use_pdserving=False, use_space_char=True, use_tensorrt=False, vis_font_path='./doc/fonts/simfang.ttf', warmup=True)
/bin/sh: 1: nvidia-smi: not found
[2021/12/27 06:53:53] root WARNING: GPU is not found in current device by nvidia-smi. Please check your device or ignore it if run on jeston.
/bin/sh: 1: nvidia-smi: not found
[2021/12/27 06:53:57] root WARNING: GPU is not found in current device by nvidia-smi. Please check your device or ignore it if run on jeston.
/bin/sh: 1: nvidia-smi: not found
[2021/12/27 06:53:57] root WARNING: GPU is not found in current device by nvidia-smi. Please check your device or ignore it if run on jeston.
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/paddleocr.py", line 439, in __init__
    super().__init__(params)
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/ppstructure/predict_system.py", line 54, in __init__
    self.table_layout = lp.PaddleDetectionLayoutModel(config_path=config_path,
  File "/usr/local/lib/python3.8/dist-packages/layoutparser/file_utils.py", line 226, in __getattr__
    raise AttributeError(f"module {self.__name__} has no attribute {name}")
AttributeError: module layoutparser has no attribute PaddleDetectionLayoutModel

Solution

Do not install the latest version of layoutparser lib from pip. Use these commands below:

wget https://paddleocr.bj.bcebos.com/whl/layoutparser-0.0.0-py3-none-any.whl
pip install -U layoutparser-0.0.0-py3-none-any.whl

ImportError: /usr/local/lib/python3.8/dist-packages/skimage/_shared/../../scikit_image.libs/libgomp-d22c30c5.so.1.0.0: cannot allocate memory in static TLS block

Error output

>>> import layoutparser as lp
/usr/lib/python3/dist-packages/apport/report.py:13: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
  import fnmatch, glob, traceback, errno, sys, atexit, locale, imp, stat
Traceback (most recent call last):
  File "/usr/local/lib/python3.8/dist-packages/skimage/__init__.py", line 151, in <module>
    from ._shared import geometry
ImportError: /usr/local/lib/python3.8/dist-packages/skimage/_shared/../../scikit_image.libs/libgomp-d22c30c5.so.1.0.0: cannot allocate memory in static TLS block

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.8/dist-packages/layoutparser/__init__.py", line 13, in <module>
    from .ocr import (
  File "/usr/local/lib/python3.8/dist-packages/layoutparser/ocr.py", line 14, in <module>
    from paddleocr import PaddleOCR
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/__init__.py", line 15, in <module>
    from .paddleocr import *
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/paddleocr.py", line 29, in <module>
    from tools.infer import predict_system
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/tools/infer/predict_system.py", line 31, in <module>
    import tools.infer.predict_rec as predict_rec
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/tools/infer/predict_rec.py", line 31, in <module>
    from ppocr.postprocess import build_post_process
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/postprocess/__init__.py", line 30, in <module>
    from .pg_postprocess import PGPostProcess
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/postprocess/pg_postprocess.py", line 25, in <module>
    from ppocr.utils.e2e_utils.pgnet_pp_utils import PGNet_PostProcess
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/utils/e2e_utils/pgnet_pp_utils.py", line 25, in <module>
    from extract_textpoint_slow import *
  File "/usr/local/lib/python3.8/dist-packages/paddleocr/ppocr/utils/e2e_utils/extract_textpoint_slow.py", line 24, in <module>
    from skimage.morphology._skeletonize import thin
  File "/usr/local/lib/python3.8/dist-packages/skimage/__init__.py", line 154, in <module>
    _raise_build_error(e)
  File "/usr/local/lib/python3.8/dist-packages/skimage/__init__.py", line 131, in _raise_build_error
    raise ImportError("""%s
ImportError: /usr/local/lib/python3.8/dist-packages/skimage/_shared/../../scikit_image.libs/libgomp-d22c30c5.so.1.0.0: cannot allocate memory in static TLS block
It seems that scikit-image has not been built correctly.

Your install of scikit-image appears to be broken.
Try re-installing the package following the instructions at:
https://scikit-image.org/docs/stable/install.html 

Solution

export LD_PRELOAD=/usr/lib/aarch64-linux-gnu/libgomp.so.1

ERROR: Cannot uninstall 'scikit-learn'. It is a distutils installed project ...

Error output

ERROR: Cannot uninstall 'scikit-learn'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall.

Solution

Use the argument with -I / --ignore-installed

pip3 install scikit-learn -I

GCC version too high If it is better to use v7.5

Solution

apt update
apt install -y build-essential
apt-get install -y manpages-dev
apt install gcc-7 g++-7
update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-7 70 --slave /usr/bin/g++ g++ /usr/bin/g++-7 --slave /usr/bin/gcov gcov /usr/bin/gcov-7

If it does not change the auto mode version, you can use sudo update-alternatives --config gcc to check and set it.

Test PaddleOCR

To get the example image first.

wget https://i.redd.it/fs8tzf72r4zz.png
mv fs8tzf72r4zz.png testimage.png

Test the package layoutparser first. Reference: README.md#2-quick-start

import cv2
import layoutparser as lp
image = cv2.imread("testimage.png")
image = image[..., ::-1]

# load model
model = lp.PaddleDetectionLayoutModel(config_path="lp://PubLayNet/ppyolov2_r50vd_dcn_365e_publaynet/config",
                                threshold=0.5,
                                label_map={0: "Text", 1: "Title", 2: "List", 3:"Table", 4:"Figure"},
                                enforce_cpu=False,
                                enable_mkldnn=True)
# detect
layout = model.detect(image)

# show result
show_img = lp.draw_box(image, layout, box_width=3, show_element_type=True)
show_img.show()

Here is the result from the sample above.

Test PaddleOCR (v2.3.0.2) Donwload the font from here We can get the simfang.ttf from the font folder.

Reference: quickstart_en.md#222-layout-analysis

import os
import cv2
from paddleocr import PPStructure,draw_structure_result,save_structure_res

table_engine = PPStructure(show_log=True)

save_folder = './'
img_path = './testimage.png'
img = cv2.imread(img_path)
result = table_engine(img)
save_structure_res(result, save_folder,os.path.basename(img_path).split('.')[0])

for line in result:
    line.pop('img')
    print(line)

from PIL import Image

font_path = './simfang.ttf'
image = Image.open(img_path).convert('RGB')
im_show = draw_structure_result(image, result,font_path=font_path)
im_show = Image.fromarray(im_show)
im_show.save('result.jpg')

Here is the result from the sample above.

ENV​

Steps​

Install Paddle​

Troubleshooting:​

Error output​

Solution:​

Error message:​

Solution​

Error output​

Solution​

Test PaddleOCR​

ENV

Steps

Install Paddle

Troubleshooting:

Error output

Solution:

Error message:

Solution

Error output

Solution

Test PaddleOCR