[技术]

【飞凌嵌入式OKMX8MP-C 开发板试用体验】tensorflow-lite-2.6编译&运行label_image例程

donatello1996

2022-2-21 00:26:04 5638

0 本帖最后由 donatello1996 于 2022-4-18 23:35 编辑【飞凌嵌入式OKMX8MP-C 开发板试用体验】tensorflow-lite-2.6编译&运行label_image例程凌厂家提供的使用手册中有提到基于tensorflow-lite的图像物品识别例程，名为label_iamge，这个例程是tensorflow-lite（下称tflite） github源码上已经有的例程，厂家所使用的tflite版本为r2.3，r表示release分支 https://github.com/tensorflow/tensorflow/tree/r2.6 复制代码下载之后，进入路径tensorflow-r2.6/tensorflow/lite/tools/make/，找到文件build_rpi_lib.sh，这个就是我们需要进行编译用的脚本，里面集成了make指令：该文件会调用tensorflow/lite/tools/make/Makefile文件，若直接运行，会提示报错： /bin/bash: ARM-linux-gnueabihf-g++: command not found 复制代码原因是该目录默认使用交叉编译方式进行编译，而如果下载到开发板上进行本地编译，需要修改编译器名称，在Makefile第310行到312行：然后是将90行的neon-vfpv4编译选项去掉，将targets目录下的rpi_makefile.inc里面的neon-vfpv4编译选项和-march=armv7-a编译选项也去掉：运行路径tensorflow-r2.6/tensorflow/lite/tools/make/下的downloads脚本： ./download_dependencies.sh 复制代码这些downloads文件夹下的下载内容都是编译tensorflow-lite-2.6 label_image例程所必需的：修改tensorflow/lite/tools/make/Makefile文件，添加label_image例程编译选项：添加编译所需要的源文件：修改tensorflow/lite/nnapi/sl/SupportLibrary.cc文件： /* Copyright 2021 The TensorFlow Authors. All Rights Reserved. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. ==============================================================================/ // Changed when importing from AOSP #include "tensorflow/lite/nnapi/sl/include/SupportLibrary.h" // Changed when importing from AOSP #include #include #include #include #include #include "tensorflow/lite/minimal_logging.h" #include "tensorflow/lite/nnapi/NeuralNetworksTypes.h" namespace tflite { namespace nnapi { using tflite::TFLITE_LOG_ERROR; NnApiSupportLibrary::NnApiSupportLibrary(const NnApiSLDriverImplFL5& impl, void libHandle) : NnApiSLDriverImplFL5(impl), libHandle(libHandle) { base.implFeatureLevel = ANEURALNETWORKS_FEATURE_LEVEL_5; } NnApiSupportLibrary::~NnApiSupportLibrary() { if (libHandle != nullptr) { dlclose(libHandle); libHandle = nullptr; } } std::unique_ptr loadNnApiSupportLibrary( const std::string& libName) { void* libHandle = dlopen(libName.c_str(), RTLD_LAZY \| RTLD_LOCAL); if (libHandle == nullptr) { TFLITE_LOG(TFLITE_LOG_ERROR, "nnapi error: unable to open library %s: %s", libName.c_str(), dlerror()); return nullptr; } auto result = loadNnApiSupportLibrary(libHandle); if (!result) { dlclose(libHandle); } return result; } template std::unique_ptr make_unique(Args&&... args) { return std::unique_ptr(new T(std::forward(args)...)); } std::unique_ptr loadNnApiSupportLibrary( void* libHandle) { NnApiSLDriverImpl* (getSlDriverImpl)(); getSlDriverImpl = reinterpret_cast( dlsym(libHandle, "ANeuralNetworks_getSLDriverImpl")); if (getSlDriverImpl == nullptr) { TFLITE_LOG(TFLITE_LOG_ERROR, "Failed to find ANeuralNetworks_getSLDriverImpl symbol"); return nullptr; } NnApiSLDriverImpl impl = getSlDriverImpl(); if (impl == nullptr) { TFLITE_LOG(TFLITE_LOG_ERROR, "ANeuralNetworks_getSLDriverImpl returned nullptr"); return nullptr; } if (impl->implFeatureLevel < ANEURALNETWORKS_FEATURE_LEVEL_5) { int64_t impl_feature_level = impl->implFeatureLevel; TFLITE_LOG(TFLITE_LOG_ERROR, "Unsupported NnApiSLDriverImpl->implFeatureLevel: %" PRId64, impl_feature_level); return nullptr; } return make_unique( *reinterpret_cast(impl), libHandle); } } // namespace nnapi } // namespace tflite 然后运行tensorflow-r2.6/tensorflow/lite/tools/make/build_rpi_lib.sh进行编译，生成目标bin文件：运行：成功。 nnapi模式下运行不成功，原因未知，等待后续。 0
2022-2-21 00:26:04　　评论淘帖0 举报相关推荐 • 【飞凌嵌入式OKMX8MP-C 开发板试用体验】SCTP通信 2877 • 【飞凌嵌入式OKMX8MP-C 开发板试用体验】出现的问题及现象 2411 • 【飞凌嵌入式OKMX8MP-C 开发板试用体验】移植官方NPU tensorflow例程并测试(运行在CPU上) 4455 • 【飞凌嵌入式OKMX8MP-C 开发板试用体验】板子已收到 2370 • 【飞凌嵌入式OKMX8MP-C 开发板试用体验】基于DPDK和ODP的OFP通信 3592 • 飞凌嵌入式OKMX8MM-C开发板简介 6112 • 飞凌嵌入式基于NXP i.MX8MMini 的OKMX8MM-C开发板试用体验 734 • 工程师测评｜飞凌嵌入式OKMX8MP-C开发板部分功能实测 1866 • 飞凌嵌入式OKMX6UL-C开发板介绍 5695 • 飞凌嵌入式OKMX6Q/DL-C开发板介绍 3462