IOS上tensorflow实现花朵识别

1、在Mac上使用Homebrew包管理器安装Python 3.6很简单，如果还没有安装Homebrew，可以在线安装。然后打开终端（Terminal）输入如下命令安装最新版Python：

brew install python3

 2、然后使用Python的包管理器pip来安装所需要的包，在终端输入如下命令：

pip3 install numpy
pip3 install scipy
pip3 install scikit-learn
pip3 install pandas
pip3 install tensorflow==1.3   #最新tensorflow1.7训练出来的模型在ios camera工程上有问题，不能识别模型的某些操作符号，                                  大概camera工程的tensorflow版本过低

      3、首先你得安装好Xcode 8，确定开发者目录指向你安装Xcode的位置并且已经被激活。（如果你在安装Xcode之前已经安装了Homebrew，这可能会指向错误的地址，导致TensorFlow安装失败）：

sudo xcode-select -s /Applications/Xcode.app/Contents/Developer

4、我们将使用名为bazel的工具来安装TensorFlow。先使用Homebrew安装所需要的包：

//brew cask install java  //安装最新java，现在是java10

brew install brew-cask
brew update
brew tap caskroom/versions   # 安装cask，如果已安装了cask可以省略。brew cask install caskroom/versions/java8
#brew cask install java8      # 使用cask安装其他的工具
brew install bazel           #要求java8
brew install automake
brew install libtool

5、下载tensorflow 1.3源码：

sudo git clone https://github.com/tensorflow/tensorflow -b r1.3

   6、下载花朵图片，在tensorflow目录下执行：

cd ..sudo mkdir tf_files
sudo curl -O http://download.tensorflow.org/example_images/flower_photos.tgz
sudo tar xzf flower_photos.tgzcd ../tensorflow

7、训练mobilenet模型：

yp@ubuntu:/home/tensorflow1.3/tensorflow$ more buildretrain.sh python3 tensorflow/examples/image_retraining/retrain.py \
--bottleneck_dir=../tf_files/bottlenecks/ \
--how_many_training_steps 20 \
--model_dir=../tf_files/inception \
--output_graph=../tf_files/retrained_graph.pb \
--output_labels=../tf_files/retrained_labels.txt \
--image_dir=../tf_files/flower_photos \
--architecture=mobilenet_1.0_224

sudo ./buildretrain.sh 

测试mobilenet模型：

yp@ubuntu:/home/tensorflow1.3/tensorflow$ more testlabel.py 
python3 tensorflow/examples/label_image/label_image.py \
--graph=../tf_files/retrained_graph.pb \
--labels=../tf_files/retrained_labels.txt \
--image=../tf_files/flower_photos/roses/99383371_37a5ac12a3_n.jpg \
--input_layer=input \
--output_layer=final_result \
--input_mean=128 \
--input_std=128 \
--input_width=224 \
--input_height=224

sudo ./testlabel.py

8、在ios camera工程中加载模型：

参考上一篇文章：https://mp.csdn.net/postedit/79838931

sudo cp ../tf_files/retrained_graph.pb ./tensorflow/examples/ios/camera/data/sudo cp ../tf_files/retrained_labels.txt ./tensorflow/examples/ios/camera/data/

修改：ios/camera/CameraExampleViewController.mm

static NSString* model_file_name = @"retrained_graph";static NSString* model_file_type = @"pb";
// This controls whether we'll be loading a plain GraphDef proto, or a
// file created by the convert_graphdef_memmapped_format utility that wraps a
// GraphDef and parameter file that can be mapped into memory from file to
// reduce overall memory usage.
const bool model_uses_memory_mapping = false;
// If you have your own model, point this to the labels file.
static NSString* labels_file_name = @"retrained_labels";
static NSString* labels_file_type = @"txt";
// These dimensions need to match those the model was trained with.
const int wanted_input_width = 224;
const int wanted_input_height = 224;
const int wanted_input_channels = 3;
const float input_mean = 117.0f;
const float input_std = 1.0f;
const std::string input_layer_name = "input";
const std::string output_layer_name = "final_result";

9、最后: 运行真机测试