https://github.com/lackhole/cutemodel-v2
Tensorflow Lite C++ Guide & Wrapper
https://github.com/lackhole/cutemodel-v2
Last synced: 2 months ago
JSON representation
Tensorflow Lite C++ Guide & Wrapper
- Host: GitHub
- URL: https://github.com/lackhole/cutemodel-v2
- Owner: lackhole
- Created: 2020-07-24T09:46:55.000Z (almost 5 years ago)
- Default Branch: master
- Last Pushed: 2021-08-29T20:47:39.000Z (almost 4 years ago)
- Last Synced: 2025-02-16T11:15:15.679Z (4 months ago)
- Size: 49.8 KB
- Stars: 5
- Watchers: 1
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Fix in progress
# C++ Wrapper of Prebuilt 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
git submodule update --init --recursive
```
## Building from source
1. Get tensorflow
```
> git clone https://github.com/tensorflow/tensorflow
```
2. Download dependencies
```
> cd tensorflow
> ./tensorflow/lite/tools/make/download_dependencies.sh
```
3. Set configuration
```
> ./configure
```
* Follow it's instruction
4. Install appropriate version of [bazel](https://docs.bazel.build/versions/master/install.html)
5. Build Libraries
* [Platform config](https://github.com/tensorflow/tensorflow/blob/master/.bazelrc#L91)
* 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
```
6. Link built library and headers to [CuteModel](https://github.com/lackhole/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
)
```
7. Link `tflite` to your main CMake project.