├── .gitignore ├── LICENSE ├── README.md ├── cpp ├── ai.props ├── libunet │ ├── ai.cpp │ ├── ai.h │ ├── cvbind.cpp │ ├── cvbind.h │ ├── libcv.vcxitems │ ├── libunet.vcxitems │ ├── unet.cpp │ ├── unet.h │ ├── unetbind.cpp │ └── unetbind.h ├── pycv │ ├── dllmain.cpp │ ├── pycv.vcxproj │ ├── pycv.vcxproj.filters │ ├── pycv.vcxproj.user │ ├── result.png │ └── test_pycv.py ├── thirdparty ├── unet.sln └── unet │ ├── main.cpp │ ├── result.png │ ├── test.png │ ├── unet.mapping │ ├── unet.vcxproj │ ├── unet.vcxproj.filters │ ├── unet.vcxproj.user │ └── unet.xml ├── doc ├── 0_pred.png ├── 0_test.png ├── u-net-architecture.png ├── unet-openvino.png ├── unet-pycv.png └── unet_20200229.png └── py3 ├── LICENSE ├── data.py ├── dataset └── membrane │ ├── test │ ├── 0.png │ ├── 0res.png │ ├── 1.png │ ├── 10.png │ ├── 11.png │ ├── 12.png │ ├── 13.png │ ├── 14.png │ ├── 15.png │ ├── 16.png │ ├── 17.png │ ├── 18.png │ ├── 19.png │ ├── 2.png │ ├── 20.png │ ├── 21.png │ ├── 22.png │ ├── 23.png │ ├── 24.png │ ├── 25.png │ ├── 26.png │ ├── 27.png │ ├── 28.png │ ├── 29.png │ ├── 3.png │ ├── 4.png │ ├── 5.png │ ├── 6.png │ ├── 7.png │ ├── 8.png │ └── 9.png │ └── train │ ├── image │ ├── 0.png │ ├── 1.png │ ├── 10.png │ ├── 11.png │ ├── 12.png │ ├── 13.png │ ├── 14.png │ ├── 15.png │ ├── 16.png │ ├── 17.png │ ├── 18.png │ ├── 19.png │ ├── 2.png │ ├── 20.png │ ├── 21.png │ ├── 22.png │ ├── 23.png │ ├── 24.png │ ├── 25.png │ ├── 26.png │ ├── 27.png │ ├── 28.png │ ├── 29.png │ ├── 3.png │ ├── 4.png │ ├── 5.png │ ├── 6.png │ ├── 7.png │ ├── 8.png │ └── 9.png │ └── label │ ├── 0.png │ ├── 1.png │ ├── 10.png │ ├── 11.png │ ├── 12.png │ ├── 13.png │ ├── 14.png │ ├── 15.png │ ├── 16.png │ ├── 17.png │ ├── 18.png │ ├── 19.png │ ├── 2.png │ ├── 20.png │ ├── 21.png │ ├── 22.png │ ├── 23.png │ ├── 24.png │ ├── 25.png │ ├── 26.png │ ├── 27.png │ ├── 28.png │ ├── 29.png │ ├── 3.png │ ├── 4.png │ ├── 5.png │ ├── 6.png │ ├── 7.png │ ├── 8.png │ └── 9.png ├── demo.py ├── model.py ├── models ├── myunet_64_1.2.h5 ├── myunet_64_1.2.pb ├── myunet_64_1.2.tflite ├── unet.mapping └── unet.xml └── utils.py /.gitignore: -------------------------------------------------------------------------------- 1 | # Cifar data 2 | *.bin 3 | *.vscode 4 | *.tar* 5 | *.tif* 6 | models/* 7 | dataset/* 8 | _*/ 9 | .vs/ 10 | x64/* 11 | cpp/x64/* 12 | thirdparty*/* -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2020 Knight 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. 22 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # UNet 网络训练和推理 2 | 3 | Train and test Unet in keras, deploy in different DL framework. 4 | 5 | 使用 Keras 训练和测试 UNet，并在深度学习推理框架下部署。 6 | 7 | ![doc/unet_20200229.png](doc/unet_20200229.png) 8 | 9 | ## 简介 10 | 11 | * 借鉴 UNet 网络原理，搭建 U 形编解码(Encoder-Decoder)语义分割网络，实现细胞分割。 12 | 13 | * 使用 Keras 训练好网络后，得到 h5 模型， 并进行推理测试。 14 | 15 | * 使用 Keras 和 tensorflow 将 h5 模型转化为 pb 模型，并进行推理测试。 16 | 17 | * 使用 OpenVino 2020.R1 对 pb 模型进行优化，得到 IR 中间文件，然后分别利用 OpenVino 的 Python 版 和 C++ 版接口进行推理。 18 | 19 | * 使用 pybind11 封装 C++ 版的推理接口，供 Python 调用。 20 | 21 | * 【WARN】更改至 tf2.x，将 h5 模型转换为 tflite 模型并推理。（慢！） 22 | 23 | ### 关于网络 24 | 25 | 原始的 UNet 网络结构如图 ： 26 | 27 | ![doc/u-net-architecture.png](doc/u-net-architecture.png) 28 | 29 | 基于 UNet 网络原理，搭建类似 UNet 的 `U` 形 Encoder-Decoder 编解码网络，实现细胞分割。修改了网络输入大小，基本块的构建及通道个数等。 30 | 31 | 修改后网络输入和输出都是: `shape=(n, 256, 256, 1), dtype=float32, range in [0.0, 1.1]`； 32 | 33 | 使用 BCE(binary-cross-entropy) 二值交叉熵作为损失函数进行训练。 34 | 35 | ### 训练和测试 36 | 37 | 数据集来自 [isbi challenge](http://brainiac2.mit.edu/isbi_challenge/)。训练图片包含 30 张 512x512x1 的图像，使用 ImageDataGenerator 扩增数据量。 38 | 39 | 20 个 epochs 后， 网络精度达到 0.96。 40 | 41 | 网络测试结果如图，(语义)分割结果看上去还可以。 42 | 43 | ![doc/0_test.png](doc/0_test.png) 44 | ![doc/0_pred.png](doc/0_pred.png) 45 | 46 | ## 使用 47 | 48 | ### 训练和测试环境 49 | 50 | * Python 3.7 51 | * Keras 2.2.4 + Tensorflow 1.14.0 (注意 keras 需要和 tensorflow 匹配) 52 | * Keras 2.3.1 + Tensorflow 2.1.0 (仅当 h5转tflite, 或 tflite 推理时用 TF 2.x， 其他的用 1.x) 53 | * OpenVino 2020.R1 54 | * pybind11 55 | 56 | ### Python 版 57 | 58 | 进入 py3 目录，分别进行训练(train)、测试(test)、保存 pb(savepb)、测试 pb(testpb)。 59 | 60 | Include `train`, `test`, `savepb`, `testpb`. 61 | 62 | ```bash 63 | cd py3 64 | ## TF1.x 65 | python demo.py --op train -d 0 66 | python demo.py --op test 67 | python demo.py --op savepb 68 | python demo.py --op testpb 69 | 70 | ## TF2.x 71 | python demo.py --op savelite 72 | python demo.py --op testlite 73 | 74 | ``` 75 | 76 | ### C++ OpenVino 推理 77 | 78 | 使用 OpenVino 进行推理的基本步骤： 79 | 80 | - 首先安装 g++/VS2015+ 和 OpenVino 工具包。 81 | - 将先前训练好的 h5 模型转为 pb 模型，然后使用 mo.py 优化 pb 模型得到 IR 文件(.xml/.bin/.mapping)。 82 | - 创建 C++ 项目，加载 IR 文件，并进行推理。 83 | 84 | 这里有 VS 项目供参考，使用方法： 85 | 86 | - 进入 cpp 文件夹 87 | - 从 [ausk/thirdparty](https://github.com/ausk/thirdparty) 下载依赖的 OpenVino 和 OpenCV 库， 解压 thirdparty 目录到 cpp 目录中。 88 | - 打开 unet.sln 解决方案文件，修改 pycv 项目中 Python 的路径。 89 | - 编译并运行 unet，直接使用 OpenVino C++ 版本推理。 90 | 91 | ![doc/unet-openvino.png](doc/unet-openvino.png) 92 | 93 | - 编译并运行 pycv，则提供 pybind11 绑定的 OpenCV。test_pycv.py 是测试文件。 94 | 95 | ![doc/unet-pycv.png](doc/unet-pycv.png) 96 | 97 | ## 参考 98 | 99 | * [U-Net: Convolutional Networks for Biomedical Image Segmentation](http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/) 100 | * keras 版 Unet 实现 [zhixuhao/unet](https://github.com/zhixuhao/unet) 101 | * OpenVino 开源版 [opencv/dldt](https://github.com/opencv/dldt) 102 | * 我的知乎文章 [191123 使用 Pybind11 和 OpenCV 创建 Python 库](https://zhuanlan.zhihu.com/p/93299698) 103 | -------------------------------------------------------------------------------- /cpp/ai.props: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | D:\Programs\Python\Python37 6 | 7 | 8 | $(SolutionDir)$(Platform)\$(ProjectName)\$(Configuration)\ 9 | 10 | 11 | 12 | $(PYTHONDIR)\include;$(SolutionDir)\thirdparty\openvino2020r1\include;$(SolutionDir)\thirdparty\opencv420\include;$(SolutionDir)\thirdparty\pybind11;%(AdditionalIncludeDirectories) 13 | WIN32;_WINDOWS;_SCL_SECURE_NO_WARNINGS;NOMINMAX;USE_OPENCV;_CRT_SECURE_NO_WARNINGS;GFLAGS_IS_A_DLL=0;_UNICODE_;UNICODE_;%(PreprocessorDefinitions) 14 | 15 | 16 | $(PYTHONDIR)\libs;$(SolutionDir)\thirdparty\openvino2020r1\lib;$(SolutionDir)\thirdparty\opencv420\lib;%(AdditionalLibraryDirectories) 17 | inference_engine.lib;opencv_core420.lib;opencv_highgui420.lib;opencv_imgproc420.lib;opencv_imgcodecs420.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies) 18 | 19 | 20 | 21 | 22 | $(PYTHONDIR) 23 | true 24 | 25 | 26 | -------------------------------------------------------------------------------- /cpp/libunet/ai.cpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/libunet/ai.cpp -------------------------------------------------------------------------------- /cpp/libunet/ai.h: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/libunet/ai.h -------------------------------------------------------------------------------- /cpp/libunet/cvbind.cpp: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2019.11.23 2 | // Modified by ausk @ 2020.03.01 3 | 4 | #include "cvbind.h" 5 | 6 | // 基本的加法 7 | int add(int i, int j) { 8 | return i + j; 9 | } 10 | 11 | // 输出 hello 12 | void sayhello() { 13 | printf("Hello!\n"); 14 | } 15 | 16 | //! 点相加 17 | cv::Point addpt(cv::Point& lhs, cv::Point& rhs) { 18 | return cv::Point(lhs.x + rhs.x, lhs.y + rhs.y); 19 | } 20 | 21 | //! 矩形相加 22 | cv::Rect addrect(cv::Rect& lhs, cv::Rect& rhs) { 23 | int x = std::min(lhs.x, rhs.x); 24 | int y = std::min(lhs.y, rhs.y); 25 | int width = std::max(lhs.x + lhs.width - 1, rhs.x + rhs.width - 1) - x + 1; 26 | int height = std::max(lhs.y + lhs.height - 1, rhs.y + rhs.height - 1) - y + 1; 27 | return cv::Rect(x, y, width, height); 28 | } 29 | 30 | //! 矩阵相加 31 | cv::Mat addmat(cv::Mat& lhs, cv::Mat& rhs) { 32 | return lhs + rhs; 33 | } 34 | 35 | //！读取图片 36 | cv::Mat imread(std::string fpath) { 37 | return cv::imread(fpath); 38 | } 39 | 40 | //! 写入图片 41 | void imwrite(std::string fpath, const cv::Mat& img) { 42 | cv::imwrite(fpath, img); 43 | } 44 | 45 | // 分离模块 46 | // https://pybind11.readthedocs.io/en/master/faq.html#how-can-i-reduce-the-build-time 47 | void bind_CV(py::module& m) { 48 | // Python + OpenCV 49 | m.def("add", &add, "A function which adds two numbers"); 50 | m.def("sayhello", &sayhello, "Just sayhello!"); 51 | m.def("addpt", &addpt, "add two point"); 52 | m.def("addrect", &addrect, "add two rect"); 53 | m.def("addmat", &addmat, "add two matrix"); 54 | m.def("imread", &imread, "read the file into np.ndarray/cv::Mat"); 55 | m.def("imwrite", &imwrite, "write np.ndarray/cv::Mat into the file"); 56 | } 57 | 58 | /* 59 | PYBIND11_MODULE(pycv, m) { 60 | m.doc() = "pycv: A small OpenCV binding created using Pybind11"; // optional module docstring 61 | bind_CV(m); 62 | }; 63 | */ -------------------------------------------------------------------------------- /cpp/libunet/cvbind.h: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2019.11.23 2 | // Modified by ausk @ 2020.03.01 3 | 4 | #pragma once 5 | 6 | #include 7 | #include 8 | #include 9 | #include 10 | #include 11 | #include 12 | 13 | namespace py = pybind11; 14 | 15 | // 绑定 cv 模块 16 | void bind_CV(py::module& m); 17 | 18 | // 对 cv::Point, cv::Rest, cv:: 19 | namespace pybind11 { 20 | namespace detail { 21 | 22 | //! (1) 点与元组转换 cv::Point <=> tuple(x,y) 23 | #if 0 24 | template<> 25 | struct type_caster { 26 | PYBIND11_TYPE_CASTER(cv::Point, _("tuple_xy")); 27 | bool load(handle obj, bool) { 28 | py::tuple pt = reinterpret_borrow(obj); 29 | value = cv::Point(pt[0].cast(), pt[1].cast()); 30 | return true; 31 | } 32 | 33 | static handle cast(const cv::Point& pt, return_value_policy, handle) { 34 | return py::make_tuple(pt.x, pt.y).release(); 35 | } 36 | }; 37 | 38 | #else 39 | template <> 40 | struct type_caster { 41 | 42 | //! 定义 cv::Point 类型名为 tuple_xy, 并声明类型为 cv::Point 的局部变量 value。 43 | PYBIND11_TYPE_CASTER(cv::Point, _("tuple_xy")); 44 | 45 | //! 步骤1：从 Python 转换到 C++。 46 | //! 将 Python tuple 对象转换为 C++ cv::Point 类型, 转换失败则返回 false。 47 | //! 其中参数2表示是否隐式类型转换。 48 | bool load(handle obj, bool) { 49 | // 确保传参是 tuple 类型 50 | if (!py::isinstance(obj)) { 51 | std::logic_error("Point(x,y) should be a tuple!"); 52 | return false; 53 | } 54 | 55 | // 从 handle 提取 tuple 对象，确保其长度是2。 56 | py::tuple pt = reinterpret_borrow(obj); 57 | if (pt.size() != 2) { 58 | std::logic_error("Point(x,y) tuple should be size of 2"); 59 | return false; 60 | } 61 | 62 | //! 将长度为2的 tuple 转换为 cv::Point。 63 | value = cv::Point(pt[0].cast(), pt[1].cast()); 64 | return true; 65 | } 66 | 67 | //! 步骤2： 从 C++ 转换到 Python。 68 | //! 将 C++ cv::Mat 对象转换到 tuple，参数2和参数3常忽略。 69 | static handle cast(const cv::Point& pt, return_value_policy, handle) { 70 | return py::make_tuple(pt.x, pt.y).release(); 71 | } 72 | }; 73 | #endif 74 | 75 | // (2) 矩形与元组转换 cv::Rect <=> tuple(x,y,w,h) 76 | template<> 77 | struct type_caster { 78 | 79 | PYBIND11_TYPE_CASTER(cv::Rect, _("tuple_xywh")); 80 | 81 | bool load(handle obj, bool) { 82 | if (!py::isinstance(obj)) { 83 | std::logic_error("Rect should be a tuple!"); 84 | return false; 85 | } 86 | 87 | py::tuple rect = reinterpret_borrow(obj); 88 | if (rect.size() != 4) { 89 | std::logic_error("Rect (x,y,w,h) tuple should be size of 4"); 90 | return false; 91 | } 92 | 93 | value = cv::Rect(rect[0].cast(), rect[1].cast(), rect[2].cast(), rect[3].cast()); 94 | return true; 95 | } 96 | 97 | static handle cast(const cv::Rect& rect, return_value_policy, handle) { 98 | return py::make_tuple(rect.x, rect.y, rect.width, rect.height).release(); 99 | } 100 | }; 101 | 102 | // (3) cv::Mat 与 numpy.ndarray 之间转换 103 | // Python 支持一种通用的插件间数据交换缓冲区协议(buffer protocol)。 104 | // 让类型暴露一个缓冲区视图(buffer view), 这样可以对原始内部数据进行直接访问，常用于矩阵类型。 105 | // 106 | //Pybind11 提供了 py::buffer_info 类型，来映射 Python 缓冲区协议(buffer protocol)。 107 | // 108 | //struct buffer_info { 109 | // void* ptr; /* Pointer to buffer */ 110 | // ssize_t itemsize; /* Size of one scalar */ 111 | // std::string format; /* Python struct-style format descriptor */ 112 | // ssize_t ndim; /* Number of dimensions */ 113 | // std::vector shape; /* Buffer dimensions */ 114 | // std::vector strides; /* Strides (in bytes) for each index */ 115 | //}; 116 | 117 | template<> 118 | struct type_caster { 119 | public: 120 | 121 | PYBIND11_TYPE_CASTER(cv::Mat, _("numpy.ndarray")); 122 | 123 | //! 1. cast numpy.ndarray to cv::Mat 124 | bool load(handle obj, bool) { 125 | array b = reinterpret_borrow(obj); 126 | buffer_info info = b.request(); 127 | 128 | int nh = 1; 129 | int nw = 1; 130 | int nc = 1; 131 | int ndims = info.ndim; 132 | if (ndims == 2) { 133 | nh = info.shape[0]; 134 | nw = info.shape[1]; 135 | } 136 | else if (ndims == 3) { 137 | nh = info.shape[0]; 138 | nw = info.shape[1]; 139 | nc = info.shape[2]; 140 | } 141 | else { 142 | throw std::logic_error("Only support 2d, 2d matrix"); 143 | return false; 144 | } 145 | 146 | int dtype; 147 | if (info.format == format_descriptor::format()) { 148 | dtype = CV_8UC(nc); 149 | } 150 | else if (info.format == format_descriptor::format()) { 151 | dtype = CV_32SC(nc); 152 | } 153 | else if (info.format == format_descriptor::format()) { 154 | dtype = CV_32FC(nc); 155 | } 156 | else { 157 | throw std::logic_error("Unsupported type, only support uchar, int32, float"); 158 | return false; 159 | } 160 | value = cv::Mat(nh, nw, dtype, info.ptr); 161 | return true; 162 | } 163 | 164 | //! 2. cast cv::Mat to numpy.ndarray 165 | static handle cast(const cv::Mat& mat, return_value_policy, handle defval) { 166 | std::string format = format_descriptor::format(); 167 | size_t elemsize = sizeof(unsigned char); 168 | int nw = mat.cols; 169 | int nh = mat.rows; 170 | int nc = mat.channels(); 171 | int depth = mat.depth(); 172 | int type = mat.type(); 173 | int dim = (depth == type) ? 2 : 3; 174 | if (depth == CV_8U) { 175 | format = format_descriptor::format(); 176 | elemsize = sizeof(unsigned char); 177 | } 178 | else if (depth == CV_32S) { 179 | format = format_descriptor::format(); 180 | elemsize = sizeof(int); 181 | } 182 | else if (depth == CV_32F) { 183 | format = format_descriptor::format(); 184 | elemsize = sizeof(float); 185 | } 186 | else { 187 | throw std::logic_error("Unsupport type, only support uchar, int32, float"); 188 | } 189 | 190 | std::vector bufferdim; 191 | std::vector strides; 192 | if (dim == 2) { 193 | bufferdim = { (size_t)nh, (size_t)nw }; 194 | strides = { elemsize * (size_t)nw, elemsize }; 195 | } 196 | else if (dim == 3) { 197 | bufferdim = { (size_t)nh, (size_t)nw, (size_t)nc }; 198 | strides = { (size_t)elemsize * nw * nc, (size_t)elemsize * nc, (size_t)elemsize }; 199 | } 200 | return array(buffer_info(mat.data, elemsize, format, dim, bufferdim, strides)).release(); 201 | } 202 | }; 203 | } 204 | }//! 205 | -------------------------------------------------------------------------------- /cpp/libunet/libcv.vcxitems: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | $(MSBuildAllProjects);$(MSBuildThisFileFullPath) 5 | true 6 | {cedae11d-b646-4a10-8fc0-9d68a54864b7} 7 | 8 | 9 | 10 | %(AdditionalIncludeDirectories);$(MSBuildThisFileDirectory) 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | -------------------------------------------------------------------------------- /cpp/libunet/libunet.vcxitems: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | $(MSBuildAllProjects);$(MSBuildThisFileFullPath) 5 | true 6 | {46a879d4-4ac4-4fc0-9307-76caf8b8af5c} 7 | 8 | 9 | 10 | %(AdditionalIncludeDirectories);$(MSBuildThisFileDirectory) 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | -------------------------------------------------------------------------------- /cpp/libunet/unet.cpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/libunet/unet.cpp -------------------------------------------------------------------------------- /cpp/libunet/unet.h: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2020.03.01 2 | 3 | #pragma once 4 | #include "ai.h" 5 | 6 | class UNetInfer { 7 | IAI* punet; 8 | std::string model_xmlfpath; 9 | 10 | public: 11 | 12 | UNetInfer(const std::string& xmlfpath){ 13 | model_xmlfpath = xmlfpath; 14 | punet = nullptr; 15 | init_UNet(); 16 | } 17 | 18 | ~UNetInfer() { 19 | destroy_UNet(); 20 | } 21 | 22 | cv::Mat operator()(const cv::Mat& src) { 23 | return run_UNet(src); 24 | } 25 | 26 | void init_UNet(); 27 | void destroy_UNet(); 28 | cv::Mat run_UNet(const cv::Mat&); 29 | }; 30 | -------------------------------------------------------------------------------- /cpp/libunet/unetbind.cpp: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2020.03.01 2 | 3 | #include "unetbind.h" 4 | 5 | void bind_UNet(py::module& m){ 6 | py::class_(m, "UNetInfer") 7 | .def(py::init()) 8 | .def("__call__", &UNetInfer::operator()) 9 | .def("init_UNet", &UNetInfer::init_UNet) 10 | .def("destroy_UNet", &UNetInfer::destroy_UNet); 11 | } -------------------------------------------------------------------------------- /cpp/libunet/unetbind.h: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2020.03.01 2 | 3 | #pragma once 4 | #include "cvbind.h" 5 | #include "unet.h" 6 | 7 | namespace py = pybind11; 8 | 9 | void bind_UNet(py::module& m); 10 | -------------------------------------------------------------------------------- /cpp/pycv/dllmain.cpp: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2019.11.23 2 | // Modified by ausk @ 2020.03.01 3 | 4 | #include "cvbind.h" 5 | #include "unetbind.h" 6 | 7 | namespace py = pybind11; 8 | 9 | PYBIND11_MODULE(pycv, m) { 10 | m.doc() = "pycv: A binding created using pybind11"; 11 | bind_CV(m); 12 | bind_UNet(m); 13 | }; 14 | -------------------------------------------------------------------------------- /cpp/pycv/pycv.vcxproj: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | Debug 6 | x64 7 | 8 | 9 | Release 10 | x64 11 | 12 | 13 | 14 | 16.0 15 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32} 16 | Win32Proj 17 | pycv 18 | 10.0 19 | pycv 20 | 21 | 22 | 23 | DynamicLibrary 24 | true 25 | v142 26 | Unicode 27 | 28 | 29 | DynamicLibrary 30 | false 31 | v142 32 | true 33 | MultiByte 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | false 53 | .pyd 54 | D:\Programs\Python\Python37\include;$(VC_IncludePath);$(WindowsSDK_IncludePath); 55 | 56 | 57 | true 58 | 59 | 60 | 61 | NotUsing 62 | Level3 63 | true 64 | true 65 | false 66 | NDEBUG;PYCV_EXPORTS;_WINDOWS;_USRDLL;%(PreprocessorDefinitions) 67 | true 68 | pch.h 69 | 70 | 71 | Windows 72 | true 73 | true 74 | true 75 | false 76 | D:\Programs\Python\Python37\libs;$(SolutionDir)\thirdparty\openvino2020r1\lib;$(SolutionDir)\thirdparty\opencv420\lib;%(AdditionalLibraryDirectories) 77 | python37.lib;inference_engine.lib;opencv_core420.lib;opencv_highgui420.lib;opencv_imgproc420.lib;opencv_imgcodecs420.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies) 78 | 79 | 80 | 81 | 82 | Use 83 | Level3 84 | true 85 | _DEBUG;PYCV_EXPORTS;_WINDOWS;_USRDLL;%(PreprocessorDefinitions) 86 | true 87 | pch.h 88 | 89 | 90 | Windows 91 | true 92 | false 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | -------------------------------------------------------------------------------- /cpp/pycv/pycv.vcxproj.filters: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | 5 | {4FC737F1-C7A5-4376-A066-2A32D752A2FF} 6 | cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx 7 | 8 | 9 | {93995380-89BD-4b04-88EB-625FBE52EBFB} 10 | h;hh;hpp;hxx;hm;inl;inc;ipp;xsd 11 | 12 | 13 | {67DA6AB6-F800-4c08-8B7A-83BB121AAD01} 14 | rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav;mfcribbon-ms 15 | 16 | 17 | 18 | 19 | 源文件 20 | 21 | 22 | -------------------------------------------------------------------------------- /cpp/pycv/pycv.vcxproj.user: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | -------------------------------------------------------------------------------- /cpp/pycv/result.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/pycv/result.png -------------------------------------------------------------------------------- /cpp/pycv/test_pycv.py: -------------------------------------------------------------------------------- 1 | # 2019/11/23 初始版本 2 | # 2020/03/01 增加 unet 推理功能 3 | # Created by ausk @ 2019.11.23 4 | # Modified by ausk @ 2020.03.01 5 | 6 | # (1) https://pybind11.readthedocs.io/en/master/basics.html 7 | # (2) [191123 使用 Pybind11 和 OpenCV 创建 Python 库](https://zhuanlan.zhihu.com/p/93299698) 8 | 9 | import os, sys 10 | curdir = os.path.dirname(os.path.abspath(__file__)) 11 | 12 | sys.path.append(os.path.join(curdir, "../x64/Release")) 13 | import pycv 14 | import numpy as np 15 | 16 | # 测试绑定的 OpenCV 17 | def test_cv(): 18 | print("\n===== Test pybind11 for OpenCV =====") 19 | # 矩阵相加 20 | a = np.random.random((2,3)).astype(np.float32) 21 | b = np.ones((2,3)).astype(np.float32) 22 | c = pycv.addmat(a, b) 23 | print("a = \n", a) 24 | print("b = \n", b) 25 | print("c = \n", c) 26 | 27 | # 点相加 28 | a = (1,2) 29 | b = (3,-1) 30 | c = pycv.addpt(a, b) 31 | print("{} + {} = {}".format(a, b, c)) 32 | 33 | # 矩形相加 34 | a = (10, 20, 20, 10) # lt(10, 20) => rb(30, 30) 35 | b = (5, 5, 20, 20) # lt(5, 5) => rb(25, 25) 36 | c = pycv.addrect(a, b) # lt(5, 5) => rb(30, 30) => wh(25, 25) 37 | print("a = \n", a) 38 | print("b = \n", b) 39 | print("c = \n", c) 40 | print("===== Finish =====\n") 41 | 42 | # 测试使用 pybind11 绑定的 OpenVino+OpenCV 实现 UNet 推理 43 | def test_unet(): 44 | print("\n===== Test pybind11 for UNet (OpenCV+OpenVino) =====") 45 | xmlfpath = "../unet/unet.xml" 46 | imgfpath = "../unet/test.png" 47 | 48 | img = pycv.imread(imgfpath) 49 | unet = pycv.UNetInfer(xmlfpath) 50 | res = unet(img) 51 | pycv.imwrite("result.png", res) 52 | print("Saved into result.png") 53 | print("===== Finish =====\n") 54 | 55 | if __name__ == "__main__": 56 | test_cv() 57 | test_unet() 58 | print("\nDone!") -------------------------------------------------------------------------------- /cpp/thirdparty: -------------------------------------------------------------------------------- 1 | D:/Programs/intel/install/thirdparty -------------------------------------------------------------------------------- /cpp/unet.sln: -------------------------------------------------------------------------------- 1 |  2 | Microsoft Visual Studio Solution File, Format Version 12.00 3 | # Visual Studio Version 16 4 | VisualStudioVersion = 16.0.29806.167 5 | MinimumVisualStudioVersion = 10.0.40219.1 6 | Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "unet", "unet\unet.vcxproj", "{8D0021F7-D7A7-47F5-8851-04EF8035533A}" 7 | EndProject 8 | Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "libunet", "libunet\libunet.vcxitems", "{46A879D4-4AC4-4FC0-9307-76CAF8B8AF5C}" 9 | EndProject 10 | Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "pycv", "pycv\pycv.vcxproj", "{673832B6-F4E7-4CD7-88DC-B5A978DEFE32}" 11 | EndProject 12 | Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "libcv", "libunet\libcv.vcxitems", "{CEDAE11D-B646-4A10-8FC0-9D68A54864B7}" 13 | EndProject 14 | Global 15 | GlobalSection(SharedMSBuildProjectFiles) = preSolution 16 | libunet\libunet.vcxitems*{46a879d4-4ac4-4fc0-9307-76caf8b8af5c}*SharedItemsImports = 9 17 | libunet\libcv.vcxitems*{673832b6-f4e7-4cd7-88dc-b5a978defe32}*SharedItemsImports = 4 18 | libunet\libunet.vcxitems*{673832b6-f4e7-4cd7-88dc-b5a978defe32}*SharedItemsImports = 4 19 | libunet\libunet.vcxitems*{8d0021f7-d7a7-47f5-8851-04ef8035533a}*SharedItemsImports = 4 20 | libunet\libcv.vcxitems*{cedae11d-b646-4a10-8fc0-9d68a54864b7}*SharedItemsImports = 9 21 | EndGlobalSection 22 | GlobalSection(SolutionConfigurationPlatforms) = preSolution 23 | Debug|x64 = Debug|x64 24 | Debug|x86 = Debug|x86 25 | Release|x64 = Release|x64 26 | Release|x86 = Release|x86 27 | EndGlobalSection 28 | GlobalSection(ProjectConfigurationPlatforms) = postSolution 29 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Debug|x64.ActiveCfg = Debug|x64 30 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Debug|x64.Build.0 = Debug|x64 31 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Debug|x86.ActiveCfg = Debug|x64 32 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Release|x64.ActiveCfg = Release|x64 33 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Release|x64.Build.0 = Release|x64 34 | {8D0021F7-D7A7-47F5-8851-04EF8035533A}.Release|x86.ActiveCfg = Release|x64 35 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Debug|x64.ActiveCfg = Debug|x64 36 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Debug|x64.Build.0 = Debug|x64 37 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Debug|x86.ActiveCfg = Debug|x64 38 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Release|x64.ActiveCfg = Release|x64 39 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Release|x64.Build.0 = Release|x64 40 | {673832B6-F4E7-4CD7-88DC-B5A978DEFE32}.Release|x86.ActiveCfg = Release|x64 41 | EndGlobalSection 42 | GlobalSection(SolutionProperties) = preSolution 43 | HideSolutionNode = FALSE 44 | EndGlobalSection 45 | GlobalSection(ExtensibilityGlobals) = postSolution 46 | SolutionGuid = {2C3C6793-7FC0-45F9-91A4-E1F641DD1500} 47 | EndGlobalSection 48 | EndGlobal 49 | -------------------------------------------------------------------------------- /cpp/unet/main.cpp: -------------------------------------------------------------------------------- 1 | // Created by ausk @ 2020.03.01 2 | 3 | #include "unet.h" 4 | 5 | int main(int argc, char* argv[]) { 6 | printf("Usage: xxx unet.xml test.png\n"); 7 | printf("Start test..."); 8 | std::string xmlfpath = "unet.xml"; 9 | std::string imgfpath = "test.png"; 10 | if(argc==3){ 11 | xmlfpath = argv[1]; 12 | imgfpath = argv[2]; 13 | } 14 | 15 | auto infer = UNetInfer(xmlfpath); 16 | cv::Mat img = cv::imread(imgfpath); 17 | cv::Mat prob = infer(img); 18 | cv::imwrite("result.png", prob); 19 | 20 | printf("Great Job!\n"); 21 | return 0; 22 | } -------------------------------------------------------------------------------- /cpp/unet/result.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/unet/result.png -------------------------------------------------------------------------------- /cpp/unet/test.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/cpp/unet/test.png -------------------------------------------------------------------------------- /cpp/unet/unet.mapping: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | 127 | 128 | 129 | 130 | 131 | 132 | 133 | 134 | 135 | 136 | 137 | 138 | 139 | 140 | 141 | 142 | 143 | 144 | 145 | 146 | 147 | 148 | 149 | 150 | 151 | 152 | 153 | 154 | 155 | 156 | 157 | 158 | 159 | 160 | 161 | 162 | 163 | 164 | 165 | 166 | 167 | 168 | 169 | 170 | 171 | 172 | 173 | 174 | 175 | 176 | 177 | 178 | 179 | 180 | 181 | 182 | 183 | 184 | 185 | 186 | 187 | 188 | 189 | 190 | 191 | 192 | 193 | 194 | 195 | 196 | 197 | 198 | 199 | 200 | 201 | 202 | 203 | 204 | 205 | 206 | 207 | 208 | 209 | 210 | 211 | 212 | 213 | 214 | 215 | 216 | 217 | 218 | 219 | 220 | 221 | 222 | 223 | 224 | 225 | 226 | 227 | 228 | 229 | 230 | 231 | 232 | 233 | 234 | 235 | 236 | 237 | 238 | 239 | 240 | 241 | 242 | 243 | 244 | 245 | 246 | 247 | 248 | 249 | 250 | 251 | 252 | 253 | 254 | 255 | 256 | 257 | 258 | 259 | 260 | 261 | 262 | 263 | 264 | 265 | 266 | 267 | 268 | 269 | 270 | 271 | 272 | 273 | 274 | 275 | 276 | 277 | 278 | 279 | 280 | 281 | 282 | 283 | 284 | 285 | 286 | 287 | 288 | 289 | 290 | 291 | 292 | 293 | 294 | 295 | 296 | 297 | 298 | 299 | 300 | 301 | 302 | 303 | 304 | 305 | 306 | 307 | 308 | 309 | 310 | 311 | 312 | 313 | 314 | 315 | 316 | 317 | 318 | 319 | 320 | 321 | 322 | 323 | 324 | 325 | 326 | 327 | 328 | 329 | 330 | 331 | 332 | 333 | 334 | 335 | 336 | 337 | 338 | 339 | 340 | 341 | 342 | 343 | 344 | 345 | 346 | 347 | 348 | 349 | 350 | 351 | 352 | 353 | 354 | 355 | 356 | 357 | 358 | 359 | 360 | -------------------------------------------------------------------------------- /cpp/unet/unet.vcxproj: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | Debug 6 | x64 7 | 8 | 9 | Release 10 | x64 11 | 12 | 13 | 14 | 15 | 16 | 17 | 16.0 18 | {8D0021F7-D7A7-47F5-8851-04EF8035533A} 19 | unet 20 | 10.0 21 | 22 | 23 | 24 | Application 25 | true 26 | v142 27 | Unicode 28 | 29 | 30 | Application 31 | false 32 | v142 33 | true 34 | MultiByte 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | true 53 | 54 | 55 | false 56 | 57 | 58 | 59 | Level3 60 | true 61 | _DEBUG;_CONSOLE;%(PreprocessorDefinitions) 62 | true 63 | 64 | 65 | Console 66 | true 67 | 68 | 69 | 70 | 71 | Level3 72 | true 73 | true 74 | true 75 | NDEBUG;_CONSOLE;%(PreprocessorDefinitions) 76 | true 77 | 78 | 79 | Console 80 | true 81 | true 82 | true 83 | 84 | 85 | 86 | 87 | 88 | -------------------------------------------------------------------------------- /cpp/unet/unet.vcxproj.filters: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | 5 | {4FC737F1-C7A5-4376-A066-2A32D752A2FF} 6 | cpp;c;cc;cxx;def;odl;idl;hpj;bat;asm;asmx 7 | 8 | 9 | {93995380-89BD-4b04-88EB-625FBE52EBFB} 10 | h;hh;hpp;hxx;hm;inl;inc;ipp;xsd 11 | 12 | 13 | {67DA6AB6-F800-4c08-8B7A-83BB121AAD01} 14 | rc;ico;cur;bmp;dlg;rc2;rct;bin;rgs;gif;jpg;jpeg;jpe;resx;tiff;tif;png;wav;mfcribbon-ms 15 | 16 | 17 | 18 | 19 | 源文件 20 | 21 | 22 | -------------------------------------------------------------------------------- /cpp/unet/unet.vcxproj.user: -------------------------------------------------------------------------------- 1 |  2 | 3 | 4 | -------------------------------------------------------------------------------- /cpp/unet/unet.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 1 9 | 1 10 | 256 11 | 256 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 64 20 | 1 21 | 3 22 | 3 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 1 31 | 1 32 | 256 33 | 256 34 | 35 | 36 | 64 37 | 1 38 | 3 39 | 3 40 | 41 | 42 | 43 | 44 | 1 45 | 64 46 | 256 47 | 256 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 1 56 | 64 57 | 1 58 | 1 59 | 60 | 61 | 62 | 63 | 64 | 65 | 1 66 | 64 67 | 256 68 | 256 69 | 70 | 71 | 1 72 | 64 73 | 1 74 | 1 75 | 76 | 77 | 78 | 79 | 1 80 | 64 81 | 256 82 | 256 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 1 91 | 64 92 | 256 93 | 256 94 | 95 | 96 | 97 | 98 | 1 99 | 64 100 | 256 101 | 256 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 1 110 | 64 111 | 256 112 | 256 113 | 114 | 115 | 116 | 117 | 1 118 | 64 119 | 128 120 | 128 121 | 122 | 123 | 124 | 125 | 126 | 127 | 128 | 77 129 | 64 130 | 3 131 | 3 132 | 133 | 134 | 135 | 136 | 137 | 138 | 139 | 1 140 | 64 141 | 128 142 | 128 143 | 144 | 145 | 77 146 | 64 147 | 3 148 | 3 149 | 150 | 151 | 152 | 153 | 1 154 | 77 155 | 128 156 | 128 157 | 158 | 159 | 160 | 161 | 162 | 163 | 164 | 1 165 | 77 166 | 1 167 | 1 168 | 169 | 170 | 171 | 172 | 173 | 174 | 1 175 | 77 176 | 128 177 | 128 178 | 179 | 180 | 1 181 | 77 182 | 1 183 | 1 184 | 185 | 186 | 187 | 188 | 1 189 | 77 190 | 128 191 | 128 192 | 193 | 194 | 195 | 196 | 197 | 198 | 199 | 1 200 | 77 201 | 128 202 | 128 203 | 204 | 205 | 206 | 207 | 1 208 | 77 209 | 128 210 | 128 211 | 212 | 213 | 214 | 215 | 216 | 217 | 218 | 1 219 | 77 220 | 128 221 | 128 222 | 223 | 224 | 225 | 226 | 1 227 | 77 228 | 64 229 | 64 230 | 231 | 232 | 233 | 234 | 235 | 236 | 237 | 92 238 | 77 239 | 3 240 | 3 241 | 242 | 243 | 244 | 245 | 246 | 247 | 248 | 1 249 | 77 250 | 64 251 | 64 252 | 253 | 254 | 92 255 | 77 256 | 3 257 | 3 258 | 259 | 260 | 261 | 262 | 1 263 | 92 264 | 64 265 | 64 266 | 267 | 268 | 269 | 270 | 271 | 272 | 273 | 1 274 | 92 275 | 1 276 | 1 277 | 278 | 279 | 280 | 281 | 282 | 283 | 1 284 | 92 285 | 64 286 | 64 287 | 288 | 289 | 1 290 | 92 291 | 1 292 | 1 293 | 294 | 295 | 296 | 297 | 1 298 | 92 299 | 64 300 | 64 301 | 302 | 303 | 304 | 305 | 306 | 307 | 308 | 1 309 | 92 310 | 64 311 | 64 312 | 313 | 314 | 315 | 316 | 1 317 | 92 318 | 64 319 | 64 320 | 321 | 322 | 323 | 324 | 325 | 326 | 327 | 1 328 | 92 329 | 64 330 | 64 331 | 332 | 333 | 334 | 335 | 1 336 | 92 337 | 32 338 | 32 339 | 340 | 341 | 342 | 343 | 344 | 345 | 346 | 111 347 | 92 348 | 3 349 | 3 350 | 351 | 352 | 353 | 354 | 355 | 356 | 357 | 1 358 | 92 359 | 32 360 | 32 361 | 362 | 363 | 111 364 | 92 365 | 3 366 | 3 367 | 368 | 369 | 370 | 371 | 1 372 | 111 373 | 32 374 | 32 375 | 376 | 377 | 378 | 379 | 380 | 381 | 382 | 1 383 | 111 384 | 1 385 | 1 386 | 387 | 388 | 389 | 390 | 391 | 392 | 1 393 | 111 394 | 32 395 | 32 396 | 397 | 398 | 1 399 | 111 400 | 1 401 | 1 402 | 403 | 404 | 405 | 406 | 1 407 | 111 408 | 32 409 | 32 410 | 411 | 412 | 413 | 414 | 415 | 416 | 417 | 1 418 | 111 419 | 32 420 | 32 421 | 422 | 423 | 424 | 425 | 1 426 | 111 427 | 32 428 | 32 429 | 430 | 431 | 432 | 433 | 434 | 435 | 436 | 1 437 | 111 438 | 32 439 | 32 440 | 441 | 442 | 443 | 444 | 1 445 | 111 446 | 16 447 | 16 448 | 449 | 450 | 451 | 452 | 453 | 454 | 455 | 133 456 | 111 457 | 3 458 | 3 459 | 460 | 461 | 462 | 463 | 464 | 465 | 466 | 1 467 | 111 468 | 16 469 | 16 470 | 471 | 472 | 133 473 | 111 474 | 3 475 | 3 476 | 477 | 478 | 479 | 480 | 1 481 | 133 482 | 16 483 | 16 484 | 485 | 486 | 487 | 488 | 489 | 490 | 491 | 1 492 | 133 493 | 1 494 | 1 495 | 496 | 497 | 498 | 499 | 500 | 501 | 1 502 | 133 503 | 16 504 | 16 505 | 506 | 507 | 1 508 | 133 509 | 1 510 | 1 511 | 512 | 513 | 514 | 515 | 1 516 | 133 517 | 16 518 | 16 519 | 520 | 521 | 522 | 523 | 524 | 525 | 526 | 1 527 | 133 528 | 16 529 | 16 530 | 531 | 532 | 533 | 534 | 1 535 | 133 536 | 16 537 | 16 538 | 539 | 540 | 541 | 542 | 543 | 544 | 545 | 2 546 | 547 | 548 | 549 | 550 | 551 | 552 | 553 | 1 554 | 133 555 | 16 556 | 16 557 | 558 | 559 | 2 560 | 561 | 562 | 563 | 564 | 1 565 | 133 566 | 32 567 | 32 568 | 569 | 570 | 571 | 572 | 573 | 574 | 575 | 1 576 | 133 577 | 32 578 | 32 579 | 580 | 581 | 1 582 | 111 583 | 32 584 | 32 585 | 586 | 587 | 588 | 589 | 1 590 | 244 591 | 32 592 | 32 593 | 594 | 595 | 596 | 597 | 598 | 599 | 600 | 133 601 | 244 602 | 3 603 | 3 604 | 605 | 606 | 607 | 608 | 609 | 610 | 611 | 1 612 | 244 613 | 32 614 | 32 615 | 616 | 617 | 133 618 | 244 619 | 3 620 | 3 621 | 622 | 623 | 624 | 625 | 1 626 | 133 627 | 32 628 | 32 629 | 630 | 631 | 632 | 633 | 634 | 635 | 636 | 1 637 | 133 638 | 1 639 | 1 640 | 641 | 642 | 643 | 644 | 645 | 646 | 1 647 | 133 648 | 32 649 | 32 650 | 651 | 652 | 1 653 | 133 654 | 1 655 | 1 656 | 657 | 658 | 659 | 660 | 1 661 | 133 662 | 32 663 | 32 664 | 665 | 666 | 667 | 668 | 669 | 670 | 671 | 1 672 | 133 673 | 32 674 | 32 675 | 676 | 677 | 678 | 679 | 1 680 | 133 681 | 32 682 | 32 683 | 684 | 685 | 686 | 687 | 688 | 689 | 690 | 2 691 | 692 | 693 | 694 | 695 | 696 | 697 | 698 | 1 699 | 133 700 | 32 701 | 32 702 | 703 | 704 | 2 705 | 706 | 707 | 708 | 709 | 1 710 | 133 711 | 64 712 | 64 713 | 714 | 715 | 716 | 717 | 718 | 719 | 720 | 1 721 | 133 722 | 64 723 | 64 724 | 725 | 726 | 1 727 | 92 728 | 64 729 | 64 730 | 731 | 732 | 733 | 734 | 1 735 | 225 736 | 64 737 | 64 738 | 739 | 740 | 741 | 742 | 743 | 744 | 745 | 111 746 | 225 747 | 3 748 | 3 749 | 750 | 751 | 752 | 753 | 754 | 755 | 756 | 1 757 | 225 758 | 64 759 | 64 760 | 761 | 762 | 111 763 | 225 764 | 3 765 | 3 766 | 767 | 768 | 769 | 770 | 1 771 | 111 772 | 64 773 | 64 774 | 775 | 776 | 777 | 778 | 779 | 780 | 781 | 1 782 | 111 783 | 1 784 | 1 785 | 786 | 787 | 788 | 789 | 790 | 791 | 1 792 | 111 793 | 64 794 | 64 795 | 796 | 797 | 1 798 | 111 799 | 1 800 | 1 801 | 802 | 803 | 804 | 805 | 1 806 | 111 807 | 64 808 | 64 809 | 810 | 811 | 812 | 813 | 814 | 815 | 816 | 1 817 | 111 818 | 64 819 | 64 820 | 821 | 822 | 823 | 824 | 1 825 | 111 826 | 64 827 | 64 828 | 829 | 830 | 831 | 832 | 833 | 834 | 835 | 2 836 | 837 | 838 | 839 | 840 | 841 | 842 | 843 | 1 844 | 111 845 | 64 846 | 64 847 | 848 | 849 | 2 850 | 851 | 852 | 853 | 854 | 1 855 | 111 856 | 128 857 | 128 858 | 859 | 860 | 861 | 862 | 863 | 864 | 865 | 1 866 | 111 867 | 128 868 | 128 869 | 870 | 871 | 1 872 | 77 873 | 128 874 | 128 875 | 876 | 877 | 878 | 879 | 1 880 | 188 881 | 128 882 | 128 883 | 884 | 885 | 886 | 887 | 888 | 889 | 890 | 92 891 | 188 892 | 3 893 | 3 894 | 895 | 896 | 897 | 898 | 899 | 900 | 901 | 1 902 | 188 903 | 128 904 | 128 905 | 906 | 907 | 92 908 | 188 909 | 3 910 | 3 911 | 912 | 913 | 914 | 915 | 1 916 | 92 917 | 128 918 | 128 919 | 920 | 921 | 922 | 923 | 924 | 925 | 926 | 1 927 | 92 928 | 1 929 | 1 930 | 931 | 932 | 933 | 934 | 935 | 936 | 1 937 | 92 938 | 128 939 | 128 940 | 941 | 942 | 1 943 | 92 944 | 1 945 | 1 946 | 947 | 948 | 949 | 950 | 1 951 | 92 952 | 128 953 | 128 954 | 955 | 956 | 957 | 958 | 959 | 960 | 961 | 1 962 | 92 963 | 128 964 | 128 965 | 966 | 967 | 968 | 969 | 1 970 | 92 971 | 128 972 | 128 973 | 974 | 975 | 976 | 977 | 978 | 979 | 980 | 2 981 | 982 | 983 | 984 | 985 | 986 | 987 | 988 | 1 989 | 92 990 | 128 991 | 128 992 | 993 | 994 | 2 995 | 996 | 997 | 998 | 999 | 1 1000 | 92 1001 | 256 1002 | 256 1003 | 1004 | 1005 | 1006 | 1007 | 1008 | 1009 | 1010 | 77 1011 | 92 1012 | 3 1013 | 3 1014 | 1015 | 1016 | 1017 | 1018 | 1019 | 1020 | 1021 | 1 1022 | 92 1023 | 256 1024 | 256 1025 | 1026 | 1027 | 77 1028 | 92 1029 | 3 1030 | 3 1031 | 1032 | 1033 | 1034 | 1035 | 1 1036 | 77 1037 | 256 1038 | 256 1039 | 1040 | 1041 | 1042 | 1043 | 1044 | 1045 | 1046 | 1 1047 | 77 1048 | 1 1049 | 1 1050 | 1051 | 1052 | 1053 | 1054 | 1055 | 1056 | 1 1057 | 77 1058 | 256 1059 | 256 1060 | 1061 | 1062 | 1 1063 | 77 1064 | 1 1065 | 1 1066 | 1067 | 1068 | 1069 | 1070 | 1 1071 | 77 1072 | 256 1073 | 256 1074 | 1075 | 1076 | 1077 | 1078 | 1079 | 1080 | 1081 | 1 1082 | 77 1083 | 256 1084 | 256 1085 | 1086 | 1087 | 1088 | 1089 | 1 1090 | 77 1091 | 256 1092 | 256 1093 | 1094 | 1095 | 1096 | 1097 | 1098 | 1099 | 1100 | 64 1101 | 77 1102 | 3 1103 | 3 1104 | 1105 | 1106 | 1107 | 1108 | 1109 | 1110 | 1111 | 1 1112 | 77 1113 | 256 1114 | 256 1115 | 1116 | 1117 | 64 1118 | 77 1119 | 3 1120 | 3 1121 | 1122 | 1123 | 1124 | 1125 | 1 1126 | 64 1127 | 256 1128 | 256 1129 | 1130 | 1131 | 1132 | 1133 | 1134 | 1135 | 1136 | 1 1137 | 64 1138 | 1 1139 | 1 1140 | 1141 | 1142 | 1143 | 1144 | 1145 | 1146 | 1 1147 | 64 1148 | 256 1149 | 256 1150 | 1151 | 1152 | 1 1153 | 64 1154 | 1 1155 | 1 1156 | 1157 | 1158 | 1159 | 1160 | 1 1161 | 64 1162 | 256 1163 | 256 1164 | 1165 | 1166 | 1167 | 1168 | 1169 | 1170 | 1171 | 1 1172 | 64 1173 | 256 1174 | 256 1175 | 1176 | 1177 | 1178 | 1179 | 1 1180 | 64 1181 | 256 1182 | 256 1183 | 1184 | 1185 | 1186 | 1187 | 1188 | 1189 | 1190 | 64 1191 | 64 1192 | 3 1193 | 3 1194 | 1195 | 1196 | 1197 | 1198 | 1199 | 1200 | 1201 | 1 1202 | 64 1203 | 256 1204 | 256 1205 | 1206 | 1207 | 64 1208 | 64 1209 | 3 1210 | 3 1211 | 1212 | 1213 | 1214 | 1215 | 1 1216 | 64 1217 | 256 1218 | 256 1219 | 1220 | 1221 | 1222 | 1223 | 1224 | 1225 | 1226 | 1 1227 | 64 1228 | 1 1229 | 1 1230 | 1231 | 1232 | 1233 | 1234 | 1235 | 1236 | 1 1237 | 64 1238 | 256 1239 | 256 1240 | 1241 | 1242 | 1 1243 | 64 1244 | 1 1245 | 1 1246 | 1247 | 1248 | 1249 | 1250 | 1 1251 | 64 1252 | 256 1253 | 256 1254 | 1255 | 1256 | 1257 | 1258 | 1259 | 1260 | 1261 | 1 1262 | 64 1263 | 256 1264 | 256 1265 | 1266 | 1267 | 1268 | 1269 | 1 1270 | 64 1271 | 256 1272 | 256 1273 | 1274 | 1275 | 1276 | 1277 | 1278 | 1279 | 1280 | 1 1281 | 64 1282 | 1 1283 | 1 1284 | 1285 | 1286 | 1287 | 1288 | 1289 | 1290 | 1291 | 1 1292 | 64 1293 | 256 1294 | 256 1295 | 1296 | 1297 | 1 1298 | 64 1299 | 1 1300 | 1 1301 | 1302 | 1303 | 1304 | 1305 | 1 1306 | 1 1307 | 256 1308 | 256 1309 | 1310 | 1311 | 1312 | 1313 | 1314 | 1315 | 1316 | 1 1317 | 1 1318 | 1 1319 | 1 1320 | 1321 | 1322 | 1323 | 1324 | 1325 | 1326 | 1 1327 | 1 1328 | 256 1329 | 256 1330 | 1331 | 1332 | 1 1333 | 1 1334 | 1 1335 | 1 1336 | 1337 | 1338 | 1339 | 1340 | 1 1341 | 1 1342 | 256 1343 | 256 1344 | 1345 | 1346 | 1347 | 1348 | 1349 | 1350 | 1 1351 | 1 1352 | 256 1353 | 256 1354 | 1355 | 1356 | 1357 | 1358 | 1 1359 | 1 1360 | 256 1361 | 256 1362 | 1363 | 1364 | 1365 | 1366 | 1367 | 1368 | 1 1369 | 1 1370 | 256 1371 | 256 1372 | 1373 | 1374 | 1375 | 1376 | 1377 | 1378 | 1379 | 1380 | 1381 | 1382 | 1383 | 1384 | 1385 | 1386 | 1387 | 1388 | 1389 | 1390 | 1391 | 1392 | 1393 | 1394 | 1395 | 1396 | 1397 | 1398 | 1399 | 1400 | 1401 | 1402 | 1403 | 1404 | 1405 | 1406 | 1407 | 1408 | 1409 | 1410 | 1411 | 1412 | 1413 | 1414 | 1415 | 1416 | 1417 | 1418 | 1419 | 1420 | 1421 | 1422 | 1423 | 1424 | 1425 | 1426 | 1427 | 1428 | 1429 | 1430 | 1431 | 1432 | 1433 | 1434 | 1435 | 1436 | 1437 | 1438 | 1439 | 1440 | 1441 | 1442 | 1443 | 1444 | 1445 | 1446 | 1447 | 1448 | 1449 | 1450 | 1451 | 1452 | 1453 | 1454 | 1455 | 1456 | 1457 | 1458 | 1459 | 1460 | 1461 | 1462 | 1463 | 1464 | 1465 | 1466 | 1467 | 1468 | 1469 | 1470 | 1471 | 1472 | 1473 | 1474 | 1475 | 1476 | 1477 | 1478 | 1479 | 1480 | 1481 | 1482 | 1483 | 1484 | 1485 | 1486 | 1487 | 1488 | 1489 | 1490 | 1491 | 1492 | 1493 | 1494 | 1495 | 1496 | 1497 | 1498 | 1499 | 1500 | 1501 | 1502 | -------------------------------------------------------------------------------- /doc/0_pred.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/0_pred.png -------------------------------------------------------------------------------- /doc/0_test.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/0_test.png -------------------------------------------------------------------------------- /doc/u-net-architecture.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/u-net-architecture.png -------------------------------------------------------------------------------- /doc/unet-openvino.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/unet-openvino.png -------------------------------------------------------------------------------- /doc/unet-pycv.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/unet-pycv.png -------------------------------------------------------------------------------- /doc/unet_20200229.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/doc/unet_20200229.png -------------------------------------------------------------------------------- /py3/LICENSE: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2020 jinlj(ausk) 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. -------------------------------------------------------------------------------- /py3/data.py: -------------------------------------------------------------------------------- 1 | # 2020/02/29 by ausk 2 | 3 | import os 4 | import cv2 5 | import numpy as np 6 | from keras.preprocessing.image import ImageDataGenerator 7 | 8 | def trainGenerator(batch_size,train_dpath,image_dname,mask_dname,aug_dict, 9 | image_color_mode = "grayscale", mask_color_mode = "grayscale", 10 | image_save_prefix = "image",mask_save_prefix = "mask", 11 | save_to_dir = None,target_size = (256,256),seed = 1): 12 | 13 | image_generator = ImageDataGenerator(**aug_dict).flow_from_directory( 14 | train_dpath, 15 | classes = [image_dname], 16 | class_mode = None, 17 | color_mode = image_color_mode, 18 | target_size = target_size, 19 | batch_size = batch_size, 20 | save_to_dir = save_to_dir, 21 | save_prefix = image_save_prefix, 22 | seed = seed) 23 | 24 | mask_generator = ImageDataGenerator(**aug_dict).flow_from_directory( 25 | train_dpath, 26 | classes = [mask_dname], 27 | class_mode = None, 28 | color_mode = mask_color_mode, 29 | target_size = target_size, 30 | batch_size = batch_size, 31 | save_to_dir = save_to_dir, 32 | save_prefix = mask_save_prefix, 33 | seed = seed) 34 | 35 | def preprocess(img,mask): 36 | if(np.max(img) > 1): 37 | img = img / 255 38 | mask = np.where(mask >127.5, 1.0, 0.0) 39 | return (img,mask) 40 | 41 | for (img,mask) in zip(image_generator, mask_generator): 42 | img,mask = preprocess(img,mask) 43 | yield (img,mask) 44 | 45 | 46 | #读图，灰度化，归一化，缩放，改变维度 [1, nh, nw, 1] 47 | def testGenerator(test_dpath, num_image = 30, input_shape = (256,256, 1)): 48 | for i in range(num_image): 49 | fpath = "{}/{}.png".format(test_dpath, i) 50 | img = cv2.imread(fpath, cv2.IMREAD_GRAYSCALE) 51 | nh, nw = img.shape[:2] 52 | if img.shape[0]>input_shape[0] or img.shape[1] > input_shape[1]: 53 | interpolation = cv2.INTER_AREA 54 | else: 55 | interpolation = cv2.INTER_CUBIC 56 | img = cv2.resize(img, (input_shape[1], input_shape[0]), interpolation = interpolation) 57 | img = np.float32(img)/255 58 | img = img[None, ..., None] 59 | yield img 60 | 61 | # 保存结果 62 | def saveResult(dpath, npyfile,num_class = 2, postfix=""): 63 | if not postfix: 64 | postfix = "_res" 65 | for i,item in enumerate(npyfile): 66 | img = np.clip(item[:,:,0]*255, 0, 255) 67 | fname = "{}{}.png".format(i, postfix) 68 | cv2.imwrite(os.path.join(dpath, fname),img) 69 | -------------------------------------------------------------------------------- /py3/dataset/membrane/test/0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/0.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/0res.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/0res.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/1.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/10.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/11.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/11.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/12.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/12.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/13.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/13.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/14.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/15.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/15.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/16.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/16.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/17.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/17.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/18.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/18.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/19.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/19.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/2.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/20.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/20.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/21.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/21.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/22.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/22.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/23.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/23.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/24.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/24.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/25.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/25.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/26.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/26.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/27.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/27.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/28.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/28.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/29.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/29.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/3.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/3.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/4.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/4.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/5.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/5.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/6.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/6.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/7.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/8.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/8.png -------------------------------------------------------------------------------- /py3/dataset/membrane/test/9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/test/9.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/0.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/1.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/10.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/11.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/11.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/12.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/12.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/13.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/13.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/14.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/15.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/15.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/16.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/16.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/17.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/17.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/18.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/18.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/19.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/19.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/2.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/20.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/20.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/21.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/21.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/22.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/22.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/23.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/23.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/24.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/24.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/25.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/25.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/26.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/26.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/27.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/27.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/28.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/28.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/29.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/29.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/3.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/3.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/4.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/4.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/5.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/5.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/6.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/6.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/7.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/8.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/8.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/image/9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/image/9.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/0.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/1.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/10.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/11.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/11.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/12.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/12.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/13.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/13.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/14.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/15.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/15.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/16.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/16.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/17.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/17.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/18.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/18.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/19.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/19.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/2.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/20.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/20.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/21.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/21.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/22.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/22.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/23.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/23.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/24.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/24.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/25.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/25.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/26.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/26.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/27.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/27.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/28.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/28.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/29.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/29.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/3.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/3.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/4.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/4.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/5.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/5.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/6.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/6.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/7.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/8.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/8.png -------------------------------------------------------------------------------- /py3/dataset/membrane/train/label/9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/dataset/membrane/train/label/9.png -------------------------------------------------------------------------------- /py3/demo.py: -------------------------------------------------------------------------------- 1 | # 2020/02/29 by ausk 2 | 3 | import warnings 4 | warnings.filterwarnings("ignore") 5 | 6 | import os, sys 7 | import argparse 8 | 9 | import numpy as np 10 | import cv2 11 | 12 | import tensorflow as tf 13 | from tensorflow.keras import backend as K 14 | #from keras.losses import * 15 | from tensorflow.keras.optimizers import Adam 16 | from tensorflow.keras import models 17 | from tensorflow.keras.callbacks import ModelCheckpoint, LearningRateScheduler 18 | 19 | from data import trainGenerator,testGenerator, saveResult 20 | from model import myunet, UNetH5, UNetPB, UNetTfLite 21 | import utils 22 | 23 | 24 | class cfg: 25 | losses = { 26 | #"dice": dice_coef_loss, 27 | "bce": "binary_crossentropy", 28 | } 29 | loss_name = "bce" 30 | 31 | custom_objects = {} # {"dice_coef_loss": dice_coef_loss} 32 | train_dpath = "dataset/membrane/train/" 33 | test_dpath = "dataset/membrane/test/" 34 | 35 | model_dpath = "models/{}".format(loss_name) 36 | pretrained_fpath = "{}/myunet_final.h5".format(model_dpath) 37 | mfpath = model_dpath + '/ep{epoch:03d}-loss{loss:.3f}-acc{accuracy:.3f}.h5' 38 | 39 | input_size = (256,256,1) 40 | epochs = 20 41 | tag = "myunet" 42 | prefix = "{}_".format(tag) if tag else "" 43 | postfix = "_{}".format(tag) if tag else "" 44 | 45 | @classmethod 46 | def setparam(cls, tag, loss_name, lr=1E-4): 47 | cls.tag=tag 48 | cls.prefix = "{}_".format(tag) if tag else "" 49 | cls.postfix = "_{}".format(tag) if tag else "" 50 | cls.loss_name = loss_name 51 | cls.loss = cls.losses.get(loss_name) 52 | cls.model_dpath = "models/{}{}".format(cls.prefix, loss_name) 53 | utils.mksured(cls.model_dpath) 54 | cls.pretrained_fpath = "{}/myunet_final.h5".format(cls.model_dpath) 55 | cls.mfpath = cls.model_dpath + '/ep{epoch:03d}-loss{loss:.3f}-acc{accuracy:.3f}.h5' 56 | cls.lr=lr 57 | 58 | @classmethod 59 | def get_lastepoch(cls): 60 | try: 61 | fpaths = glob.glob(cls.model_dpath+"/ep*.h5") 62 | print(fpaths) 63 | last_epoch = sorted(int(fpath.split("/")[-1][2:5]) for fpath in fpaths)[-1] 64 | except: 65 | last_epoch = 0 66 | return last_epoch 67 | 68 | 69 | # 训练 UNet 70 | def train_unet(cfg, resume=True): 71 | input_size = cfg.input_size 72 | pretrained_fpath = cfg.pretrained_fpath 73 | model_dpath = cfg.model_dpath 74 | loss = cfg.loss 75 | train_dpath = cfg.train_dpath 76 | lr = cfg.lr 77 | epochs = cfg.epochs 78 | mfpath = cfg.mfpath #model_dpath + '/ep{epoch:03d}-loss{loss:.3f}-acc{accuracy:.3f}.h5' 79 | last_epoch = cfg.get_lastepoch() 80 | print(last_epoch) 81 | 82 | # (1) 创建训练生成器 83 | augdict = dict(rotation_range=0.2, 84 | width_shift_range=0.05, 85 | height_shift_range=0.05, 86 | shear_range=0.05, 87 | zoom_range=0.05, 88 | horizontal_flip=True, 89 | fill_mode='nearest') 90 | train_gen = trainGenerator(4, train_dpath,'image','label',augdict, save_to_dir = None) 91 | 92 | if False: 93 | for i,batch in enumerate(train_gen): 94 | if i >= 3: 95 | break 96 | 97 | # (2) 编译和训练模型, (可选地加载预训练模型，继续训练） 98 | model = myunet(input_size) 99 | model.compile(optimizer = Adam(lr =lr), loss = loss, metrics = ['accuracy']) 100 | #model.summary() 101 | 102 | # 加载预训练模型 103 | if resume and (pretrained_fpath) and os.path.exists(pretrained_fpath): 104 | try: 105 | model.load_weights(pretrained_fpath) 106 | except Exception as exc: 107 | print("Exception: {}".format(exc)) 108 | 109 | # 训练模型 110 | model_checkpoint = ModelCheckpoint(mfpath, monitor='loss', verbose=1, save_best_only=True) 111 | model_tb = keras.callbacks.TensorBoard(log_dir=model_dpath, histogram_freq=0, write_graph=False, write_images=False) 112 | model.fit_generator(train_gen, steps_per_epoch=1000, epochs=epochs, callbacks=[model_checkpoint, model_tb], initial_epoch=last_epoch) 113 | model.save(pretrained_fpath) 114 | 115 | # 测试 UNet 116 | def test_unet( cfg ): 117 | pretrained_fpath = cfg.pretrained_fpath 118 | test_dpath = cfg.test_dpath 119 | postfix = cfg.postfix 120 | 121 | tf.keras.backend.set_learning_phase(0) 122 | K.set_learning_phase(0) 123 | 124 | # 从文件中加载模型 125 | print("Loading: {}".format(pretrained_fpath)) 126 | #model = models.load_model(pretrained_fpath, custom_objects=custom_objects) 127 | model = UNetH5(pretrained_fpath, (256, 256, 1)) 128 | 129 | # 创建数据生成器，预测并保存结果 130 | test_gen = testGenerator(test_dpath) 131 | results = model.predict_generator(test_gen,30,verbose=1) 132 | saveResult(test_dpath,results, postfix=postfix) 133 | 134 | 135 | # 保存为 PB 文件 136 | def run_savepb(h5fpath, pbfpath=None): 137 | if pbfpath is None: 138 | pbfpath = h5fpath[:-2] + "pb" 139 | #K.clear_session() 140 | #K.set_session(tf.Session()) 141 | print(h5fpath) 142 | sess = tf.Session() 143 | with sess.as_default(): 144 | model = models.load_model(h5fpath) 145 | #model.load_weights(h5fpath, by_name=True) 146 | utils.savepb(model, sess, pbname = pbfpath) 147 | 148 | def test_loadpb(pbfpath): 149 | unet = UNetPB(pbfpath, (256, 256, 1)) 150 | for i in range(30): 151 | fpath = "dataset/membrane/test/{}.png".format(i) 152 | fpath2 = "dataset/membrane/test/{}_pb.png".format(i) 153 | img = cv2.imread(fpath) 154 | print(fpath, img.shape) 155 | gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) 156 | 157 | with utils.timeit(): 158 | result = unet.predict(gray) 159 | 160 | print(result.min(), result.max(), result.shape) 161 | cv2.imwrite(fpath2, result) 162 | 163 | def run_savelite(h5fpath, litefpath=None): 164 | if litefpath is None: 165 | litefpath = h5fpath[:-2] + "lite" 166 | utils.keras2tflite(h5fpath, litefpath) 167 | 168 | def test_loadlite(litefpath): 169 | unet = UNetTfLite(litefpath) 170 | for i in range(30): 171 | fpath = "dataset/membrane/test/{}.png".format(i) 172 | fpath2 = "dataset/membrane/test/{}_pb.png".format(i) 173 | img = cv2.imread(fpath) 174 | print(fpath, img.shape) 175 | gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) 176 | 177 | with utils.timeit(): 178 | result = unet.predict(gray) 179 | 180 | 181 | def get_args(): 182 | #tag = "myunet" 183 | parser = argparse.ArgumentParser(description="Train/Test Unet-like CNN") 184 | parser.add_argument("--devices", "-d", default="0", type=str) 185 | parser.add_argument("--op", default="test", type=str) 186 | 187 | args = parser.parse_args(sys.argv[1:]) 188 | assert args.op in ("train", "test", "savepb", "testpb", "savelite", "testlite") 189 | return args 190 | 191 | 192 | if __name__ == "__main__": 193 | args = get_args() 194 | devices = args.devices 195 | op = args.op 196 | 197 | # (0) 设置参数 198 | #assert tag in ("myunet", ) 199 | os.environ["CUDA_VISIBLE_DEVICES"] = devices 200 | cfg.setparam(tag="myunet",loss_name="bce", lr=1E-4) 201 | 202 | # (1) 训练 203 | if op == "train": 204 | train_unet(cfg) 205 | 206 | # (2) 测试 207 | if op == "test": 208 | cfg.pretrained_fpath = "models/myunet_64_1.2.h5" 209 | test_unet(cfg) 210 | 211 | # (3) 转 pb 模型 212 | if op == "savepb": 213 | h5fpath = cfg.pretrained_fpath 214 | h5fpath = "models/myunet_64_1.2.h5" 215 | pbfpath = "models/myunet_64_1.2.pb" 216 | run_savepb(h5fpath) 217 | 218 | # (4) 测试 pb 模型 219 | if op == "testpb": 220 | pbfpath = "models/myunet_64_1.2.pb" 221 | test_loadpb(pbfpath) 222 | 223 | # (5) 测试保存 lite 224 | if op == "savelite": 225 | h5fpath = "models/myunet_64_1.2.h5" 226 | run_savelite(h5fpath) 227 | 228 | # (6) 测试 tflite 229 | if op == "testlite": 230 | litefpath = "models/myunet_64_1.2.tflite" 231 | test_loadlite(litefpath) 232 | -------------------------------------------------------------------------------- /py3/model.py: -------------------------------------------------------------------------------- 1 | # 2020/02/29 by ausk 2 | 3 | import utils 4 | import warnings 5 | warnings.filterwarnings("ignore") 6 | 7 | import os, sys, time 8 | import numpy as np 9 | import cv2 10 | 11 | import tensorflow as tf 12 | from tensorflow.keras import layers, models 13 | from tensorflow.keras import backend as K 14 | 15 | # 处理块的名字 16 | def handle_block_names(prefix, stage): 17 | name = "{}_{}".format(prefix, stage) 18 | conv_name = '{}_conv'.format(name) 19 | bn_name = '{}_bn'.format(name) 20 | relu_name = '{}_relu'.format(name) 21 | return conv_name, bn_name, relu_name 22 | 23 | # 默认的 ConvReLU6 24 | def ConvReLU( nfilter, kernel_size, prefix="block", stage=1, use_bn=False): 25 | conv_name, bn_name, relu_name = handle_block_names(prefix, stage) 26 | def layer(x): 27 | conv_name, bn_name, relu_name = handle_block_names(prefix, stage) 28 | x = layers.Conv2D( nfilter, kernel_size, padding="same", kernel_initializer = 'glorot_uniform', activation = None, name=conv_name, use_bias=not(use_bn))(x) 29 | if use_bn: 30 | x = layers.BatchNormalization(epsilon=1e-3, momentum=0.999, name=bn_name)(x) 31 | #x = layers.Activation('relu', name=relu_name)(x) 32 | x = layers.ReLU(6.0, name=relu_name)(x) 33 | return x 34 | return layer 35 | 36 | # 编码模块 37 | def EncoderBlock( nfilter, stage, kernel_size=(3,3), use_bn=False, use_drop=False, use_pool = True): 38 | prefix = "encoder" 39 | def layer(x): 40 | if use_pool: 41 | x = layers.MaxPooling2D(pool_size=(2, 2))(x) 42 | x = ConvReLU( nfilter, kernel_size, use_bn=use_bn, prefix=prefix, stage = str(stage)+"1")(x) 43 | #x = ConvReLU( nfilter, kernel_size, use_bn=use_bn, prefix=prefix, stage = str(stage)+"2")(x) 44 | if use_drop: 45 | x = layers.Dropout(0.5)(x) 46 | return x 47 | return layer 48 | 49 | # 解码模块 50 | def DecoderBlock( nfilter, stage, kernel_size=(3,3), use_bn=False, use_up=True, skip=None): 51 | prefix = "decoder" 52 | def layer(x): 53 | nonlocal skip 54 | if use_up: 55 | x = layers.UpSampling2D(size=(2,2), name="{}_{}_up".format(prefix, str(stage)))(x) 56 | #x = ConvReLU( nfilter, kernel_size, use_bn=use_bn, prefix=prefix, stage = str(stage)+"0")(x) 57 | if skip is not None: 58 | x = layers.Concatenate()([x, skip]) 59 | x = ConvReLU( nfilter, kernel_size, use_bn=use_bn, prefix=prefix, stage = str(stage)+"1")(x) 60 | #x = ConvReLU( nfilter, kernel_size, use_bn=use_bn, prefix=prefix, stage = str(stage)+"2")(x) 61 | return x 62 | return layer 63 | 64 | # 搭建 UNet 65 | def myunet(input_size=(256, 256, 1), nfilter0=64,ratio=1.2): 66 | inputs = layers.Input(input_size) 67 | stages = 5 68 | nfilters = [int(nfilter0 * (ratio**i) + 0.5) for i in range(stages)] 69 | feats = [] 70 | x = inputs 71 | for stage in range(1, stages+1): 72 | nfilter = nfilters[stage-1] 73 | use_drop = stage > 4 74 | use_pool = stage > 1 75 | use_bn = True #stage > 1 76 | x = EncoderBlock( nfilter, stage, kernel_size=(3,3), use_bn=use_bn, use_drop=use_drop, use_pool=use_pool)(x) 77 | feats.append(x) 78 | 79 | for stage in range(stages, 0, -1): 80 | nfilter = nfilters[stage-1] 81 | use_drop = False 82 | use_bn = True #stage > 1 83 | use_up = stage > 1 84 | skip = feats[stage-2] if stage>2 else None 85 | x = DecoderBlock( nfilter, stage, kernel_size=(3,3), use_bn=use_bn, use_up=use_up, skip=skip)(x) 86 | 87 | x = ConvReLU(nfilter0, kernel_size=(3,3), prefix="final", stage=1, use_bn=True)(x) 88 | x = layers.Conv2D(1, 1, activation = 'sigmoid', name="final")(x) 89 | model = models.Model(inputs = inputs, outputs = x) 90 | model.summary() 91 | return model 92 | 93 | # UNet in H5 94 | class UNetH5: 95 | def __init__(self, h5fpath, input_shape): 96 | assert len(input_shape)==2 or input_shape[2] == 1 97 | self.input_shape = input_shape 98 | self.input_size = (int(input_shape[1]), int(input_shape[0])) 99 | tf.keras.backend.set_learning_phase(0) 100 | K.set_learning_phase(0) 101 | self.model = models.load_model(h5fpath, custom_objects={}) 102 | 103 | def predict_generator(self, test_gen, batch_size, verbose=0): 104 | return self.model.predict_generator(test_gen, batch_size, verbose=verbose) 105 | 106 | def predict(self, img): 107 | #assert img.shape == input_shape, "Make sure you have input {} ndarray" 108 | nh, nw = img.shape[:2] 109 | if img.ndim == 3 and img.shape[-1] !=1 : 110 | img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) 111 | if img.shape[:2] != self.input_shape[:2]: 112 | img = cv2.resize(img, self.input_size) 113 | ximg = (np.float32(img)/255)[None, ..., None] 114 | 115 | xout = self.model.predict(ximg)[0] #, ..., 0] 116 | xout = np.clip(xout * 255, 0, 255).astype(np.uint8) 117 | return xout 118 | 119 | # UNet in PB 120 | class UNetPB: 121 | def __init__(self, pbfpath, input_shape): 122 | self.input_shape = input_shape 123 | self.pbfpath = pbfpath 124 | self.graph = utils.loadGraph(pbfpath) 125 | self.input_node = self.graph.get_tensor_by_name("input_1:0") 126 | self.output_node = self.graph.get_tensor_by_name("output_0:0") 127 | self.session = tf.Session(graph=self.graph) 128 | 129 | def predict(self, img): 130 | nh,nw = self.input_shape[:2] 131 | img = cv2.resize(img, (nw, nh)) 132 | 133 | if img.ndim ==2: 134 | ximg = img[None, ..., None] 135 | else: 136 | ximg = img[None, ...] 137 | ximg = np.float32(ximg)/255.0 138 | 139 | xout = self.session.run(self.output_node, feed_dict={self.input_node: ximg})[0] 140 | xout = np.clip(xout * 255, 0, 255).astype(np.uint8) 141 | return xout 142 | 143 | class UNetTfLite: 144 | def __init__(self, litefpath): 145 | self.interpreter = tf.lite.Interpreter(litefpath) 146 | self.interpreter.allocate_tensors() 147 | 148 | def predict(self, img=None): 149 | input_details = self.interpreter.get_input_details() 150 | output_details = self.interpreter.get_output_details() 151 | 152 | _, nh, nw, _ = input_details[0]['shape'] 153 | img = cv2.resize(img, (nw, nh)) 154 | 155 | if img.ndim ==2: 156 | ximg = img[None, ..., None] 157 | else: 158 | ximg = img[None, ...] 159 | ximg = np.float32(ximg)/255.0 160 | 161 | self.interpreter.set_tensor(input_details[0]["index"], ximg) 162 | self.interpreter.invoke() 163 | xout = self.interpreter.get_tensor(output_details[0]["index"])[0,:,:,0] 164 | 165 | xout = np.clip(xout * 255, 0, 255).astype(np.uint8) 166 | return xout 167 | 168 | 169 | def __call__(self, *args, **kwargs): 170 | return self.infer(*args,**kwargs) 171 | 172 | def infer(self, img, verbose=False): 173 | interpreter = self.interpreter 174 | labels = self.labels 175 | input_details = interpreter.get_input_details() 176 | output_details = interpreter.get_output_details() -------------------------------------------------------------------------------- /py3/models/myunet_64_1.2.h5: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/models/myunet_64_1.2.h5 -------------------------------------------------------------------------------- /py3/models/myunet_64_1.2.pb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/models/myunet_64_1.2.pb -------------------------------------------------------------------------------- /py3/models/myunet_64_1.2.tflite: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/ausk/keras-unet-deploy/487efc85b87b4c5dbff8b0f2cb2476d183ba2397/py3/models/myunet_64_1.2.tflite -------------------------------------------------------------------------------- /py3/models/unet.mapping: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 | 76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 | 101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 | 126 | 127 | 128 | 129 | 130 | 131 | 132 | 133 | 134 | 135 | 136 | 137 | 138 | 139 | 140 | 141 | 142 | 143 | 144 | 145 | 146 | 147 | 148 | 149 | 150 | 151 | 152 | 153 | 154 | 155 | 156 | 157 | 158 | 159 | 160 | 161 | 162 | 163 | 164 | 165 | 166 | 167 | 168 | 169 | 170 | 171 | 172 | 173 | 174 | 175 | 176 | 177 | 178 | 179 | 180 | 181 | 182 | 183 | 184 | 185 | 186 | 187 | 188 | 189 | 190 | 191 | 192 | 193 | 194 | 195 | 196 | 197 | 198 | 199 | 200 | 201 | 202 | 203 | 204 | 205 | 206 | 207 | 208 | 209 | 210 | 211 | 212 | 213 | 214 | 215 | 216 | 217 | 218 | 219 | 220 | 221 | 222 | 223 | 224 | 225 | 226 | 227 | 228 | 229 | 230 | 231 | 232 | 233 | 234 | 235 | 236 | 237 | 238 | 239 | 240 | 241 | 242 | 243 | 244 | 245 | 246 | 247 | 248 | 249 | 250 | 251 | 252 | 253 | 254 | 255 | 256 | 257 | 258 | 259 | 260 | 261 | 262 | 263 | 264 | 265 | 266 | 267 | 268 | 269 | 270 | 271 | 272 | 273 | 274 | 275 | 276 | 277 | 278 | 279 | 280 | 281 | 282 | 283 | 284 | 285 | 286 | 287 | 288 | 289 | 290 | 291 | 292 | 293 | 294 | 295 | 296 | 297 | 298 | 299 | 300 | 301 | 302 | 303 | 304 | 305 | 306 | 307 | 308 | 309 | 310 | 311 | 312 | 313 | 314 | 315 | 316 | 317 | 318 | 319 | 320 | 321 | 322 | 323 | 324 | 325 | 326 | 327 | 328 | 329 | 330 | 331 | 332 | 333 | 334 | 335 | 336 | 337 | 338 | 339 | 340 | 341 | 342 | 343 | 344 | 345 | 346 | 347 | 348 | 349 | 350 | 351 | 352 | 353 | 354 | 355 | 356 | 357 | 358 | 359 | 360 | -------------------------------------------------------------------------------- /py3/models/unet.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 1 9 | 1 10 | 256 11 | 256 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 64 20 | 1 21 | 3 22 | 3 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 1 31 | 1 32 | 256 33 | 256 34 | 35 | 36 | 64 37 | 1 38 | 3 39 | 3 40 | 41 | 42 | 43 | 44 | 1 45 | 64 46 | 256 47 | 256 48 | 49 | 50 | 51 | 52 | 53 | 54 | 55 | 1 56 | 64 57 | 1 58 | 1 59 | 60 | 61 | 62 | 63 | 64 | 65 | 1 66 | 64 67 | 256 68 | 256 69 | 70 | 71 | 1 72 | 64 73 | 1 74 | 1 75 | 76 | 77 | 78 | 79 | 1 80 | 64 81 | 256 82 | 256 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 1 91 | 64 92 | 256 93 | 256 94 | 95 | 96 | 97 | 98 | 1 99 | 64 100 | 256 101 | 256 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 1 110 | 64 111 | 256 112 | 256 113 | 114 | 115 | 116 | 117 | 1 118 | 64 119 | 128 120 | 128 121 | 122 | 123 | 124 | 125 | 126 | 127 | 128 | 77 129 | 64 130 | 3 131 | 3 132 | 133 | 134 | 135 | 136 | 137 | 138 | 139 | 1 140 | 64 141 | 128 142 | 128 143 | 144 | 145 | 77 146 | 64 147 | 3 148 | 3 149 | 150 | 151 | 152 | 153 | 1 154 | 77 155 | 128 156 | 128 157 | 158 | 159 | 160 | 161 | 162 | 163 | 164 | 1 165 | 77 166 | 1 167 | 1 168 | 169 | 170 | 171 | 172 | 173 | 174 | 1 175 | 77 176 | 128 177 | 128 178 | 179 | 180 | 1 181 | 77 182 | 1 183 | 1 184 | 185 | 186 | 187 | 188 | 1 189 | 77 190 | 128 191 | 128 192 | 193 | 194 | 195 | 196 | 197 | 198 | 199 | 1 200 | 77 201 | 128 202 | 128 203 | 204 | 205 | 206 | 207 | 1 208 | 77 209 | 128 210 | 128 211 | 212 | 213 | 214 | 215 | 216 | 217 | 218 | 1 219 | 77 220 | 128 221 | 128 222 | 223 | 224 | 225 | 226 | 1 227 | 77 228 | 64 229 | 64 230 | 231 | 232 | 233 | 234 | 235 | 236 | 237 | 92 238 | 77 239 | 3 240 | 3 241 | 242 | 243 | 244 | 245 | 246 | 247 | 248 | 1 249 | 77 250 | 64 251 | 64 252 | 253 | 254 | 92 255 | 77 256 | 3 257 | 3 258 | 259 | 260 | 261 | 262 | 1 263 | 92 264 | 64 265 | 64 266 | 267 | 268 | 269 | 270 | 271 | 272 | 273 | 1 274 | 92 275 | 1 276 | 1 277 | 278 | 279 | 280 | 281 | 282 | 283 | 1 284 | 92 285 | 64 286 | 64 287 | 288 | 289 | 1 290 | 92 291 | 1 292 | 1 293 | 294 | 295 | 296 | 297 | 1 298 | 92 299 | 64 300 | 64 301 | 302 | 303 | 304 | 305 | 306 | 307 | 308 | 1 309 | 92 310 | 64 311 | 64 312 | 313 | 314 | 315 | 316 | 1 317 | 92 318 | 64 319 | 64 320 | 321 | 322 | 323 | 324 | 325 | 326 | 327 | 1 328 | 92 329 | 64 330 | 64 331 | 332 | 333 | 334 | 335 | 1 336 | 92 337 | 32 338 | 32 339 | 340 | 341 | 342 | 343 | 344 | 345 | 346 | 111 347 | 92 348 | 3 349 | 3 350 | 351 | 352 | 353 | 354 | 355 | 356 | 357 | 1 358 | 92 359 | 32 360 | 32 361 | 362 | 363 | 111 364 | 92 365 | 3 366 | 3 367 | 368 | 369 | 370 | 371 | 1 372 | 111 373 | 32 374 | 32 375 | 376 | 377 | 378 | 379 | 380 | 381 | 382 | 1 383 | 111 384 | 1 385 | 1 386 | 387 | 388 | 389 | 390 | 391 | 392 | 1 393 | 111 394 | 32 395 | 32 396 | 397 | 398 | 1 399 | 111 400 | 1 401 | 1 402 | 403 | 404 | 405 | 406 | 1 407 | 111 408 | 32 409 | 32 410 | 411 | 412 | 413 | 414 | 415 | 416 | 417 | 1 418 | 111 419 | 32 420 | 32 421 | 422 | 423 | 424 | 425 | 1 426 | 111 427 | 32 428 | 32 429 | 430 | 431 | 432 | 433 | 434 | 435 | 436 | 1 437 | 111 438 | 32 439 | 32 440 | 441 | 442 | 443 | 444 | 1 445 | 111 446 | 16 447 | 16 448 | 449 | 450 | 451 | 452 | 453 | 454 | 455 | 133 456 | 111 457 | 3 458 | 3 459 | 460 | 461 | 462 | 463 | 464 | 465 | 466 | 1 467 | 111 468 | 16 469 | 16 470 | 471 | 472 | 133 473 | 111 474 | 3 475 | 3 476 | 477 | 478 | 479 | 480 | 1 481 | 133 482 | 16 483 | 16 484 | 485 | 486 | 487 | 488 | 489 | 490 | 491 | 1 492 | 133 493 | 1 494 | 1 495 | 496 | 497 | 498 | 499 | 500 | 501 | 1 502 | 133 503 | 16 504 | 16 505 | 506 | 507 | 1 508 | 133 509 | 1 510 | 1 511 | 512 | 513 | 514 | 515 | 1 516 | 133 517 | 16 518 | 16 519 | 520 | 521 | 522 | 523 | 524 | 525 | 526 | 1 527 | 133 528 | 16 529 | 16 530 | 531 | 532 | 533 | 534 | 1 535 | 133 536 | 16 537 | 16 538 | 539 | 540 | 541 | 542 | 543 | 544 | 545 | 2 546 | 547 | 548 | 549 | 550 | 551 | 552 | 553 | 1 554 | 133 555 | 16 556 | 16 557 | 558 | 559 | 2 560 | 561 | 562 | 563 | 564 | 1 565 | 133 566 | 32 567 | 32 568 | 569 | 570 | 571 | 572 | 573 | 574 | 575 | 1 576 | 133 577 | 32 578 | 32 579 | 580 | 581 | 1 582 | 111 583 | 32 584 | 32 585 | 586 | 587 | 588 | 589 | 1 590 | 244 591 | 32 592 | 32 593 | 594 | 595 | 596 | 597 | 598 | 599 | 600 | 133 601 | 244 602 | 3 603 | 3 604 | 605 | 606 | 607 | 608 | 609 | 610 | 611 | 1 612 | 244 613 | 32 614 | 32 615 | 616 | 617 | 133 618 | 244 619 | 3 620 | 3 621 | 622 | 623 | 624 | 625 | 1 626 | 133 627 | 32 628 | 32 629 | 630 | 631 | 632 | 633 | 634 | 635 | 636 | 1 637 | 133 638 | 1 639 | 1 640 | 641 | 642 | 643 | 644 | 645 | 646 | 1 647 | 133 648 | 32 649 | 32 650 | 651 | 652 | 1 653 | 133 654 | 1 655 | 1 656 | 657 | 658 | 659 | 660 | 1 661 | 133 662 | 32 663 | 32 664 | 665 | 666 | 667 | 668 | 669 | 670 | 671 | 1 672 | 133 673 | 32 674 | 32 675 | 676 | 677 | 678 | 679 | 1 680 | 133 681 | 32 682 | 32 683 | 684 | 685 | 686 | 687 | 688 | 689 | 690 | 2 691 | 692 | 693 | 694 | 695 | 696 | 697 | 698 | 1 699 | 133 700 | 32 701 | 32 702 | 703 | 704 | 2 705 | 706 | 707 | 708 | 709 | 1 710 | 133 711 | 64 712 | 64 713 | 714 | 715 | 716 | 717 | 718 | 719 | 720 | 1 721 | 133 722 | 64 723 | 64 724 | 725 | 726 | 1 727 | 92 728 | 64 729 | 64 730 | 731 | 732 | 733 | 734 | 1 735 | 225 736 | 64 737 | 64 738 | 739 | 740 | 741 | 742 | 743 | 744 | 745 | 111 746 | 225 747 | 3 748 | 3 749 | 750 | 751 | 752 | 753 | 754 | 755 | 756 | 1 757 | 225 758 | 64 759 | 64 760 | 761 | 762 | 111 763 | 225 764 | 3 765 | 3 766 | 767 | 768 | 769 | 770 | 1 771 | 111 772 | 64 773 | 64 774 | 775 | 776 | 777 | 778 | 779 | 780 | 781 | 1 782 | 111 783 | 1 784 | 1 785 | 786 | 787 | 788 | 789 | 790 | 791 | 1 792 | 111 793 | 64 794 | 64 795 | 796 | 797 | 1 798 | 111 799 | 1 800 | 1 801 | 802 | 803 | 804 | 805 | 1 806 | 111 807 | 64 808 | 64 809 | 810 | 811 | 812 | 813 | 814 | 815 | 816 | 1 817 | 111 818 | 64 819 | 64 820 | 821 | 822 | 823 | 824 | 1 825 | 111 826 | 64 827 | 64 828 | 829 | 830 | 831 | 832 | 833 | 834 | 835 | 2 836 | 837 | 838 | 839 | 840 | 841 | 842 | 843 | 1 844 | 111 845 | 64 846 | 64 847 | 848 | 849 | 2 850 | 851 | 852 | 853 | 854 | 1 855 | 111 856 | 128 857 | 128 858 | 859 | 860 | 861 | 862 | 863 | 864 | 865 | 1 866 | 111 867 | 128 868 | 128 869 | 870 | 871 | 1 872 | 77 873 | 128 874 | 128 875 | 876 | 877 | 878 | 879 | 1 880 | 188 881 | 128 882 | 128 883 | 884 | 885 | 886 | 887 | 888 | 889 | 890 | 92 891 | 188 892 | 3 893 | 3 894 | 895 | 896 | 897 | 898 | 899 | 900 | 901 | 1 902 | 188 903 | 128 904 | 128 905 | 906 | 907 | 92 908 | 188 909 | 3 910 | 3 911 | 912 | 913 | 914 | 915 | 1 916 | 92 917 | 128 918 | 128 919 | 920 | 921 | 922 | 923 | 924 | 925 | 926 | 1 927 | 92 928 | 1 929 | 1 930 | 931 | 932 | 933 | 934 | 935 | 936 | 1 937 | 92 938 | 128 939 | 128 940 | 941 | 942 | 1 943 | 92 944 | 1 945 | 1 946 | 947 | 948 | 949 | 950 | 1 951 | 92 952 | 128 953 | 128 954 | 955 | 956 | 957 | 958 | 959 | 960 | 961 | 1 962 | 92 963 | 128 964 | 128 965 | 966 | 967 | 968 | 969 | 1 970 | 92 971 | 128 972 | 128 973 | 974 | 975 | 976 | 977 | 978 | 979 | 980 | 2 981 | 982 | 983 | 984 | 985 | 986 | 987 | 988 | 1 989 | 92 990 | 128 991 | 128 992 | 993 | 994 | 2 995 | 996 | 997 | 998 | 999 | 1 1000 | 92 1001 | 256 1002 | 256 1003 | 1004 | 1005 | 1006 | 1007 | 1008 | 1009 | 1010 | 77 1011 | 92 1012 | 3 1013 | 3 1014 | 1015 | 1016 | 1017 | 1018 | 1019 | 1020 | 1021 | 1 1022 | 92 1023 | 256 1024 | 256 1025 | 1026 | 1027 | 77 1028 | 92 1029 | 3 1030 | 3 1031 | 1032 | 1033 | 1034 | 1035 | 1 1036 | 77 1037 | 256 1038 | 256 1039 | 1040 | 1041 | 1042 | 1043 | 1044 | 1045 | 1046 | 1 1047 | 77 1048 | 1 1049 | 1 1050 | 1051 | 1052 | 1053 | 1054 | 1055 | 1056 | 1 1057 | 77 1058 | 256 1059 | 256 1060 | 1061 | 1062 | 1 1063 | 77 1064 | 1 1065 | 1 1066 | 1067 | 1068 | 1069 | 1070 | 1 1071 | 77 1072 | 256 1073 | 256 1074 | 1075 | 1076 | 1077 | 1078 | 1079 | 1080 | 1081 | 1 1082 | 77 1083 | 256 1084 | 256 1085 | 1086 | 1087 | 1088 | 1089 | 1 1090 | 77 1091 | 256 1092 | 256 1093 | 1094 | 1095 | 1096 | 1097 | 1098 | 1099 | 1100 | 64 1101 | 77 1102 | 3 1103 | 3 1104 | 1105 | 1106 | 1107 | 1108 | 1109 | 1110 | 1111 | 1 1112 | 77 1113 | 256 1114 | 256 1115 | 1116 | 1117 | 64 1118 | 77 1119 | 3 1120 | 3 1121 | 1122 | 1123 | 1124 | 1125 | 1 1126 | 64 1127 | 256 1128 | 256 1129 | 1130 | 1131 | 1132 | 1133 | 1134 | 1135 | 1136 | 1 1137 | 64 1138 | 1 1139 | 1 1140 | 1141 | 1142 | 1143 | 1144 | 1145 | 1146 | 1 1147 | 64 1148 | 256 1149 | 256 1150 | 1151 | 1152 | 1 1153 | 64 1154 | 1 1155 | 1 1156 | 1157 | 1158 | 1159 | 1160 | 1 1161 | 64 1162 | 256 1163 | 256 1164 | 1165 | 1166 | 1167 | 1168 | 1169 | 1170 | 1171 | 1 1172 | 64 1173 | 256 1174 | 256 1175 | 1176 | 1177 | 1178 | 1179 | 1 1180 | 64 1181 | 256 1182 | 256 1183 | 1184 | 1185 | 1186 | 1187 | 1188 | 1189 | 1190 | 64 1191 | 64 1192 | 3 1193 | 3 1194 | 1195 | 1196 | 1197 | 1198 | 1199 | 1200 | 1201 | 1 1202 | 64 1203 | 256 1204 | 256 1205 | 1206 | 1207 | 64 1208 | 64 1209 | 3 1210 | 3 1211 | 1212 | 1213 | 1214 | 1215 | 1 1216 | 64 1217 | 256 1218 | 256 1219 | 1220 | 1221 | 1222 | 1223 | 1224 | 1225 | 1226 | 1 1227 | 64 1228 | 1 1229 | 1 1230 | 1231 | 1232 | 1233 | 1234 | 1235 | 1236 | 1 1237 | 64 1238 | 256 1239 | 256 1240 | 1241 | 1242 | 1 1243 | 64 1244 | 1 1245 | 1 1246 | 1247 | 1248 | 1249 | 1250 | 1 1251 | 64 1252 | 256 1253 | 256 1254 | 1255 | 1256 | 1257 | 1258 | 1259 | 1260 | 1261 | 1 1262 | 64 1263 | 256 1264 | 256 1265 | 1266 | 1267 | 1268 | 1269 | 1 1270 | 64 1271 | 256 1272 | 256 1273 | 1274 | 1275 | 1276 | 1277 | 1278 | 1279 | 1280 | 1 1281 | 64 1282 | 1 1283 | 1 1284 | 1285 | 1286 | 1287 | 1288 | 1289 | 1290 | 1291 | 1 1292 | 64 1293 | 256 1294 | 256 1295 | 1296 | 1297 | 1 1298 | 64 1299 | 1 1300 | 1 1301 | 1302 | 1303 | 1304 | 1305 | 1 1306 | 1 1307 | 256 1308 | 256 1309 | 1310 | 1311 | 1312 | 1313 | 1314 | 1315 | 1316 | 1 1317 | 1 1318 | 1 1319 | 1 1320 | 1321 | 1322 | 1323 | 1324 | 1325 | 1326 | 1 1327 | 1 1328 | 256 1329 | 256 1330 | 1331 | 1332 | 1 1333 | 1 1334 | 1 1335 | 1 1336 | 1337 | 1338 | 1339 | 1340 | 1 1341 | 1 1342 | 256 1343 | 256 1344 | 1345 | 1346 | 1347 | 1348 | 1349 | 1350 | 1 1351 | 1 1352 | 256 1353 | 256 1354 | 1355 | 1356 | 1357 | 1358 | 1 1359 | 1 1360 | 256 1361 | 256 1362 | 1363 | 1364 | 1365 | 1366 | 1367 | 1368 | 1 1369 | 1 1370 | 256 1371 | 256 1372 | 1373 | 1374 | 1375 | 1376 | 1377 | 1378 | 1379 | 1380 | 1381 | 1382 | 1383 | 1384 | 1385 | 1386 | 1387 | 1388 | 1389 | 1390 | 1391 | 1392 | 1393 | 1394 | 1395 | 1396 | 1397 | 1398 | 1399 | 1400 | 1401 | 1402 | 1403 | 1404 | 1405 | 1406 | 1407 | 1408 | 1409 | 1410 | 1411 | 1412 | 1413 | 1414 | 1415 | 1416 | 1417 | 1418 | 1419 | 1420 | 1421 | 1422 | 1423 | 1424 | 1425 | 1426 | 1427 | 1428 | 1429 | 1430 | 1431 | 1432 | 1433 | 1434 | 1435 | 1436 | 1437 | 1438 | 1439 | 1440 | 1441 | 1442 | 1443 | 1444 | 1445 | 1446 | 1447 | 1448 | 1449 | 1450 | 1451 | 1452 | 1453 | 1454 | 1455 | 1456 | 1457 | 1458 | 1459 | 1460 | 1461 | 1462 | 1463 | 1464 | 1465 | 1466 | 1467 | 1468 | 1469 | 1470 | 1471 | 1472 | 1473 | 1474 | 1475 | 1476 | 1477 | 1478 | 1479 | 1480 | 1481 | 1482 | 1483 | 1484 | 1485 | 1486 | 1487 | 1488 | 1489 | 1490 | 1491 | 1492 | 1493 | 1494 | 1495 | 1496 | 1497 | 1498 | 1499 | 1500 | 1501 | 1502 | -------------------------------------------------------------------------------- /py3/utils.py: -------------------------------------------------------------------------------- 1 | # 2020/02/29 by ausk 2 | 3 | import warnings 4 | warnings.filterwarnings("ignore") 5 | 6 | import os, sys, glob 7 | import contextlib 8 | import time 9 | 10 | os.environ["TF_FORCE_GPU_ALLOW_GROWTH"] = "true" 11 | os.environ["TF_ENABLE_GPU_GARBAGE_COLLECTION"] = "true" 12 | os.environ["TF_CPP_MIN_LOG_LEVEL"] = "true" 13 | os.environ["CUDA_VISIBLE_DEVICES"] = "" 14 | 15 | import tensorflow as tf 16 | from tensorflow.python.framework import graph_util, graph_io 17 | from tensorflow.keras import models 18 | import tensorflow.keras.backend as K 19 | 20 | #tf.compat.v1.disable_eager_execution() 21 | 22 | #assert tf.__version__[:4] == "1.4.", "Tensorflow should be v1.4.x!" 23 | 24 | # Reset the GPU session 25 | def reset_gpu_session(devices="0", fraction=0.8): 26 | os.environ["TF_FORCE_GPU_ALLOW_GROWTH"] = "true" 27 | os.environ["TF_ENABLE_GPU_GARBAGE_COLLECTION"] = "true" 28 | os.environ["TF_CPP_MIN_LOG_LEVEL"] = "true" 29 | os.environ["CUDA_VISIBLE_DEVICES"] = devices 30 | config = tf.ConfigProto() 31 | config.gpu_options.allow_growth = True # dynamically grow the memory used on the GPU 32 | config.gpu_options.per_process_gpu_memory_fraction = fraction 33 | config.log_device_placement = False 34 | sess = tf.Session(config=config) 35 | K.set_session(sess) # set this TensorFlow session as the default session for Keras 36 | return sess 37 | 38 | # To get the version. 39 | def get_version(version): 40 | assert isinstance(version, str) 41 | import re 42 | return list(map(int, re.findall("(\d+)", version)))[:3] 43 | 44 | # To make sure the directory exist. 45 | def mksured(dpath): 46 | if not os.path.exists(dpath): 47 | os.makedirs(dpath, exist_ok=True) 48 | 49 | # The contextmanager to avoid raise except. 50 | @contextlib.contextmanager 51 | def warnOnException(): 52 | try: 53 | yield 54 | except Exception as exc: 55 | print("Exception: {}".format(exc)) 56 | finally: 57 | pass 58 | 59 | # The contextmanager to time the code. 60 | @contextlib.contextmanager 61 | def timeit(): 62 | ts = time.time() 63 | try: 64 | yield 65 | finally: 66 | pass 67 | te = time.time() 68 | print("dt: {:.3f} ms ".format( (te-ts)*1000) ) 69 | 70 | 71 | # Save the graph in current session as binary protobuf file. 72 | def savepb(model, sess=None, pbname="result.pb"): 73 | #tf.keras.backend.set_learning_phase(0) 74 | #K.set_learning_phase(0) 75 | model_outputs = model.outputs 76 | output_names = [] 77 | outputs = [] 78 | for i, output in enumerate(model_outputs): 79 | output_names.append("output_{}".format(i)) 80 | outputs.append(tf.identity(output, name = output_names[i])) 81 | 82 | input_names = [node.op.name for node in model.inputs] 83 | 84 | print("[ INFO ] inputs: ", input_names) 85 | print("[ INFO ] outputs: ", output_names) 86 | 87 | if sess is None: 88 | sess = K.get_session() 89 | 90 | graph_def = sess.graph.as_graph_def() 91 | #graphdef_inf = graph_util.remove_training_nodes(graph_def) 92 | #graphdef_frozen = graph_util.convert_variables_to_constants(sess, graphdef_inf, output_names) 93 | graphdef_frozen = graph_util.convert_variables_to_constants(sess, graph_def, output_names) 94 | graph_io.write_graph(graphdef_frozen, ".", pbname, as_text=False) 95 | print("[ INFO ] Saved pb into: ", pbname) 96 | print("[ INFO ] inputs: ", input_names) 97 | print("[ INFO ] outputs: ", output_names) 98 | 99 | 100 | def loadGraph(pbfpath, prefix=""): 101 | """ 102 | (1) About prefix. 103 | If we set name(prefix) = None，then it will add "import/" prefix to each node. 104 | And we should use name(prefix) = "" to keep none prefix. 105 | (2) It may be faild when load pb graph if it contains BN layers. 106 | """ 107 | tf.keras.backend.set_learning_phase(0) 108 | K.set_learning_phase(0) 109 | K.clear_session() 110 | 111 | # (1) Load the protobuf file from the disk and parse it to retrieve the unserialized graph_def. 112 | #with tf.gfile.GFile(pbfpath, "rb") as fin: 113 | with open(pbfpath, "rb") as fin: 114 | graph_def = tf.GraphDef() 115 | graph_def.ParseFromString(fin.read()) 116 | 117 | # (2) Import a graph_def into the current default Graph. 118 | with tf.Graph().as_default() as graph: 119 | tf.import_graph_def( 120 | graph_def, 121 | input_map=None, 122 | return_elements=None, 123 | name=prefix, 124 | op_dict=None, 125 | producer_op_list=None 126 | ) 127 | 128 | # We can verify that we can access the list of operations in the graph 129 | for op in graph.get_operations(): 130 | pass 131 | #print(op.name) 132 | 133 | return graph 134 | """ 135 | # Access the input and output nodes 136 | x = graph.get_tensor_by_name(prefix+'input_1:0') 137 | y = graph.get_tensor_by_name(prefix+'output_0:0') 138 | 139 | # Launch a Session and load the graph, feed and interence. 140 | with tf.Session(graph=graph) as sess: 141 | xinput = np.random.random((1, 256, 256,1)) 142 | pred_y = sess.run( y, feed_dict={x: xinput} ) 143 | print(pred_y) 144 | """ 145 | 146 | def keras2tflite(h5fpath, litefpath=None): 147 | if litefpath is None: 148 | litefpath = h5fpath[:-2] + "tflite" 149 | model = models.load_model(h5fpath) 150 | c = tf.lite.TFLiteConverter.from_keras_model(model) 151 | data = c.convert() 152 | with open(litefpath, "wb") as fout: 153 | fout.write(data) --------------------------------------------------------------------------------