OVERVIEW

该工程是Swift版本的官方Camera使用案例，并简单封装。

Build Library

tensorflow/contrib/makefile/build_all_ios.sh -a arm64 armv7s armv7 // 编译真机的版本

Link Binary With Libraries

在Build Phases -> Link Binary With Libraries中添加如下依赖:

• Accelerate.framework
• tensorflow/tensorflow/contrib/makefile/gen/lib/ios_ARM64/libtensorflow-core.a
• tensorflow/tensorflow/contrib/makefile/gen/protobuf_ios/lib/libprotobuf.a
• tensorflow/tensorflow/contrib/makefile/gen/protobuf_ios/lib/libprotobuf-lite.a
• tensorflow/tensorflow/contrib/makefile/downloads/nsync/builds/lipo.ios.c++11/nsync.a

Library Search Paths

在Build Settings -> Library Search Paths中添加如下路径：

• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/lib/ios_ARM64
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/protobuf_ios/lib
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/downloads/nsync/builds/lipo.ios.c++11

Header Search Paths

在Build Settings -> Header Search Paths中添加如下路径：

• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/downloads/protobuf/src
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/downloads/nsync/public
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/downloads/eigen
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/downloads

User Header Search Paths

在Build Settings -> User Header Search Paths中添加如下路径：

• $(PROJECT_DIR)
• $(PROJECT_DIR)/tensorflow
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/proto/

Other Linker Flags

在Build Settings -> Other Linker Flags中添加如下路径：

• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/protobuf_ios/lib/libprotobuf-lite.a
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/protobuf_ios/lib/libprotobuf.a
• -force_load
• $(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/lib/ios_ARM64/libtensorflow-core.a

其他

在Build Settings里，设置如下：

• Enable Bitcode： No
• Warnings / Documentation Comments: No
• Warnings / Deprecated Functions: No

TIPS

在头文件 include时出现 xxxx.pb.h not found

device_attributes.pb_text.h: No such file or directory 请确保$(PROJECT_DIR)/tensorflow/tensorflow/contrib/makefile/gen/proto/已加入Header Search Path

在头文件 include时出现 xxxx.h not found

请确保tensorflow/与.xcodeproj目录层级关系正确

运行时出现 No session factory registered for the given session options

请确保Other Linker Flags里的路径正确引入，如果有-all_load则改为'`ObjC'

Bazel安装

使用Homebrew

首先安装JDK8 + 、Homebrew

brew install bazel

Model引用

在移动端使用Tensorflow需要两个文件：xxx.pb和xxx.txt。pb文件是Model，文本文件是识别结果的键。 拿到Model后直接引用时如果出现Model加载失败的错误。则是Model格式问题。 1.是否将Variables于Model合并。 2.是否转换成了移动端可用的Model。

对于第一点。要AI组提供新的Model。 第二点 则在tensorflow/下(安装过bazel)

bazel run tensorflow/tools/graph_transforms:transform_graph --
--in_graph=tensorflow_inception_graph.pb
--out_graph=optimized_inception_graph.pb --inputs='Mul' --outputs='final_result'

将model转换成移动端可用的格式。时间长长久久...

识别

在进行识别过程中，需要传递以下几个参数

private static final int INPUT_WIDTH = 299;
private static final int INPUT_HEIGHT = 299;
private static final int IMAGE_MEAN = 128;
private static final float IMAGE_STD = 128;
private static final String INPUT_NAME = "Mul";
private static final String OUTPUT_NAME = "final_result";

private static final String MODEL_FILE = "retrained_graph_optimized.pb";
private static final String LABEL_FILE =
"label.txt";

接下来就可以正确的引用啦