您好作者，这个东西工程要怎么改才可以支持python3(自己装boost确实困难重重,.so的动态模块导入,python3会报错)

[canjie-luo]

您好作者，这个东西工程要怎么改才可以支持python3(自己装boost确实困难重重,.so的动态模块导入,python3会报错)

[RubanSeven]

我用python重写了这个代码，速度在14ms，有兴趣的可以支持下，https://github.com/RubanSeven/Text-Image-Augmentation-python

[tnkong]

这个工程比较困难的地方就是boost的安装，其他的都还好,尝试了python3环境下运行,安装boost的时候，编译之前给了参数 --with-python=/dir/../python3, 结果没有运行成功，在导入import Augment这句报错，无法导入动态module,缺少函数......如果作者有精力请推出python3版本,感激不尽

[keeofkoo]

@tnkong If you want to build it with python3, you just have to install boost for python3. Installing it using anaconda is a good option. For the import part, you gotta have a well-built Augment.so in the same directory for python to be able to import it.

[Canjie-Luo]

正如 @ @keeofkoo 所说，只需要安装在Python 3环境中安装好boost即可。这里需要注意的是，这个项目依赖于OpenCV 2，更高的版本不支持。这个项目虽然有用，但是使用起来确实麻烦，希望能够和大家合作把它完善一下。 @tnkong

[tnkong]

@Canjie-Luo @keeofkoo 问题卡在编译的地方了，我想应该是调用cmake命令这句话,　我没学过相关的知识，转行做深度学习，没碰过Ｃ语言相关的东西，所以...

从编译时返回的日志看，编译过程自动修改为python2.7的bosst支持,我的确也装了python3的boost支持包
一下是日志
cmake -D CUDA_USE_STATIC_CUDA_RUNTIME=OFF ..
-- The C compiler identification is GNU 4.8.5
-- The CXX compiler identification is GNU 5.4.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found PythonLibs: /usr/lib/x86_64-linux-gnu/libpython2.7.so (found suitable version "2.7.12", minimum required is "2.7") 　　＃　请留心这个地方
CMake Warning at /usr/share/cmake-3.5/Modules/FindBoost.cmake:725 (message):
Imported targets not available for Boost version 106700
Call Stack (most recent call first):
/usr/share/cmake-3.5/Modules/FindBoost.cmake:763 (_Boost_COMPONENT_DEPENDENCIES)
/usr/share/cmake-3.5/Modules/FindBoost.cmake:1332 (_Boost_MISSING_DEPENDENCIES)

我的直觉认为这个玩具需要修改CMakeLists.txt，　不过我没有对应的知识

[tnkong]

到后面执行make命令，都是用python2完成的，即使boost用python3安装的也不行，　应该是需要修改编译调用的程序
make
Scanning dependencies of target Augment
[ 12%] Building CXX object CMakeFiles/Augment.dir/src/imgwarp_piecewiseaffine.cpp.o
[ 25%] Building CXX object CMakeFiles/Augment.dir/src/imgwarp_mls_rigid.cpp.o
[ 37%] Building CXX object CMakeFiles/Augment.dir/src/delaunay.cpp.o
[ 50%] Building CXX object CMakeFiles/Augment.dir/src/imgwarp_mls.cpp.o
[ 62%] Building CXX object CMakeFiles/Augment.dir/src/conversion.cpp.o
In file included from /usr/include/python2.7/numpy/ndarraytypes.h:1777:0,
from /usr/include/python2.7/numpy/ndarrayobject.h:18,
from /home/tnkong/PycharmProjects/HanWangProJectPython/imageAugment/include/conversion.h:8,
from /home/tnkong/PycharmProjects/HanWangProJectPython/imageAugment/src/conversion.cpp:1:
/usr/include/python2.7/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]
#warning "Using deprecated NumPy API, disable it by "
^
[ 75%] Building CXX object CMakeFiles/Augment.dir/src/Augment.cpp.o
In file included from /usr/include/python2.7/numpy/ndarraytypes.h:1777:0,
from /usr/include/python2.7/numpy/ndarrayobject.h:18,
from /home/tnkong/PycharmProjects/HanWangProJectPython/imageAugment/include/conversion.h:8,
from /home/tnkong/PycharmProjects/HanWangProJectPython/imageAugment/src/Augment.cpp:7:
/usr/include/python2.7/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]
#warning "Using deprecated NumPy API, disable it by "
^
[ 87%] Building CXX object CMakeFiles/Augment.dir/src/imgwarp_mls_similarity.cpp.o
[100%] Linking CXX shared library Augment.so
[100%] Built target Augment

[Canjie-Luo]

好的，收到您的问题了，我回头研究一下哈，最近手上确实太多东西要完成。

[DuckJ]

请问在python3下面你们可以运行么？怎么将这个嵌入crnn训练做数据增强，有人试验过么？我在crnn训练的时候随机对原图进行三种变化之一，可是程序在docker里面都会报错Connection reset by peer，然后产生core.***的文件。这个项目下的demo.py我是可运行的。就是训练时候出错

[DuckJ]

im_Distort = Augment.GenerateDistort(im, 4)
我把上面的4变小一点，出现错误跟#3一样，Floating point exception.
错误是在训练了一些迭代之后出现的

[Canjie-Luo]

您好，您的输入原图不能够太小，否则扭动不了。另外，可以在
https://github.com/meijieru/crnn.pytorch/blob/d3a47f91691b31b8d5336e2ed5932e6cf65142f0/dataset.py#L69
这个类中嵌入。实在不行的话，您可以生成足够多的样本放在硬盘再训练也行。

[DuckJ]

请问这个原图太小指的图片高度大概多少啊，我用的就是crnn的synth90k那个数据集，图片高度基本都是30多像素。请问形变之前都得resize到相同的形状么

[Canjie-Luo]

是的，从这个数据集中拿出来的图像，不一定符合这个大小，需要先resize一下。