TensorFlow Lite C++ image classification demo
编译环境:docker
编译参考:https://tensorflow.google.cn/lite/guide/build_arm64
一:编译Tensorflowlite
- 从github上下载tensorflow源码,
- 地址:https://github.com/tensorflow/tensorflow
- 加速下载:
pip3 install -i http://mirrors.aliyun.com/pypi/simple/ --upgrade tensorflow
- 下载tensorflow依赖库
./tensorflow/lite/tools/make/download_dependencies.sh
- 修改编译环境
- 修改MakeFile中的交叉编译工具为板子上专用的(我这板子用的是 aarch64-linux-gnu-g++通用的arm64交叉编译工具),文件如下
- 编译
./tensorflow/lite/tools/make/build_aarch64_lib.sh
6. 验证
1. /home/tensorflowlite/tensorflow/tensorflow/lite/tools/make/gen/linux_aarch64/bin/minimal是集成了最精简的tensorflowlite的示例程序,copy到板子上
3. 上面的tflite可以从这里下载,存到了/tmp目录下
curl https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224.tgz | tar xzv -C /tmp
二:编译图像分类demo
- 示例代码地址如下,有编译的README,本文演示用源码编译https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/examples/label_image
- 直接make,生成labelimage可执行文件,copy到板子上
- 执行测试指令
./labelimage --tflite_model mobilenet_v1_1.0_224.tflite --labels labels.txt --image grace_hopper.bmp
- 上面的参数部分
- tflite模型,下载地址:(看命令,存到了/tmp目录下)
curl https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224.tgz | tar xzv -C /tmp
- labels.txt,下载地址:
curl https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_1.0_224_frozen.tgz | tar xzv -C /tmp mobilenet_v1_1.0_224/labels.txt
- bmp图片,在tensorflow源码的这个位置
tensorflow/lite/examples/label_image/testdata/grace_hopper.bmp
Original: https://blog.csdn.net/wyl530274554/article/details/119386075
Author: 面向对象World
Title: ARM64开发板运行Tensorflow lite图片分类demo
原创文章受到原创版权保护。转载请注明出处:https://www.johngo689.com/666266/
转载文章受原作者版权保护。转载请注明原作者出处!