Tensorflow Lite C++ API

Prebuilt Guide
Building from source

Prebuilt Guide

Select branch according to your platform.
CuteModel usage: https://github.com/visualcamp/Tensorflow-Lite/tree/cutemodel

git clone https://github.com/lackhole/Tensorflow-Lite
cd Tensorflow-Lite
git checkout <branch>
git submodule update --init --recursive

Building from source

Get tensorflow

> git clone https://github.com/tensorflow/tensorflow

Download dependencies

> cd tensorflow
> ./tensorflow/lite/tools/make/download_dependencies.sh

Set configuration
```
> ./configure
```
- Follow it's instruction
Install appropriate version of bazel

Build Libraries

Platform config
- Android arm64-v8a: --config=android_arm64
- Android armeabi-v7a: --config=android_arm
- iOS: --config=ios_fat
- No --config: detects current platform

iOS

bazel build --config=ios_fat -c opt //tensorflow/lite/experimental/ios:TensorFlowLiteC_framework

Other Platform

CPU

bazel build -c opt --cxxopt=--std=c++11 //tensorflow/lite/c:tensorflowlite_c

GPU

bazel build -c opt //tensorflow/lite/delegates/gpu:libtensorflowlite_gpu_delegate.so

NNAPI (Android Only)

bazel build -c opt --config=android_arm64 //tensorflow/lite/delegates/nnapi:nnapi_delegate
bazel build -c opt --config=android_arm64 //tensorflow/lite/nnapi:nnapi_implementation
bazel build -c opt --config=android_arm64 //tensorflow/lite/nnapi:nnapi_util

Link built library and headers to CuteModel using CMakeLists.txt

tflite
   ├─cutemodel // Wrapper
   ├─include   // Tensorflow Lite headers
   ├─lib       // Built Libraries
   └─CMakeLists.txt

CMakeLists.txt (Android Example)

cmake_minimum_required(VERSION 3.4.1)
project(tflite)

set (TFLITE_PATH ${CMAKE_CURRENT_SOURCE_DIR})
set (TFLITE_LIB_PATH "${TFLITE_PATH}/lib/${ANDROID_ABI}")
set (TFLITE_INCLUDE_PATH "${TFLITE_PATH}/include")
set (CUTEMODEL_PATH "${TFLITE_PATH}/cutemodel")

add_library(lib_tflite SHARED IMPORTED)
set_target_properties(lib_tflite PROPERTIES IMPORTED_LOCATION ${TFLITE_LIB_PATH}/libtensorflowlite_c.so)

add_library(lib_tflite_gpu SHARED IMPORTED)
set_target_properties(lib_tflite_gpu PROPERTIES IMPORTED_LOCATION ${TFLITE_LIB_PATH}/libtensorflowlite_gpu_delegate.so)

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++14")

add_library(tflite SHARED ${CUTEMODEL_PATH}/CuteModel.cpp)


target_include_directories(tflite PUBLIC
        ${TFLITE_PATH}
        ${TFLITE_INCLUDE_PATH}
        ${TFLITE_INCLUDE_PATH}/tensorflow/lite/tools/make/downloads/flatbuffers/include
        ${TFLITE_INCLUDE_PATH}/tensorflow/lite/tools/make/downloads/absl
        ${TFLITE_INCLUDE_PATH}/tensorflow/lite/tools/make/downloads/absl/absl
        )

target_link_libraries(tflite
        lib_tflite
        lib_tflite_gpu
        log
)

Link tflite to your main CMake project.

yahuuu/Tensorflow-Lite

Tensorflow Lite C++ API

Prebuilt Guide

Building from source