tensorflow_cpp

tensorflow_cpp is a header-only library that provides helpful wrappers around the TensorFlow C++ API, allowing you to easily load, inspect, and run saved models and frozen graphs in C++. The library is easy to integrate into CMake projects, but is also available as a ROS and ROS2 package.

[!IMPORTANT]
This repository is open-sourced and maintained by the Institute for Automotive Engineering (ika) at RWTH Aachen University.
Deep Learning is one of many research topics within our Vehicle Intelligence & Automated Driving domain.
If you would like to learn more about how we can support your deep learning or automated driving efforts, feel free to reach out to us!
     Timo Woopen - Manager Research Area Vehicle Intelligence & Automated Driving
     +49 241 80 23549
     timo.woopen@ika.rwth-aachen.de

If you are looking for an easy way to install the TensorFlow C++ API, we suggest that you also check out our repository libtensorflow_cc. There, we provide a pre-built library and a Docker image for easy installation and usage of the TensorFlow C++ API.

• tensorflow_cpp
  • Examples
  • Installation
  • Dependencies
  • CMake
  • ROS/ROS2
  • Testing
  • Documentation
  • Acknowledgements
  • Notice

Examples

Loading and running a single-input/single-output model

```cpp

include

include

include

include

include

// load single-input/single-output model std::string modelpath = "/PATH/TO/MODEL"; tensorflowcpp::Model model; model.loadModel(model_path);

// log model info std::cout << model.getInfoString() << std::endl;

// get input/output shape/type, if required std::vector inputshape = model.getInputShape(); tensorflow::DataType outputtype = model.getOutputType(); // ... do something ...

// create and fill input tensor tensorflow::Tensor input_tensor; // ... fill input tensor ...

// run model tensorflow::Tensor outputtensor = model(inputtensor); ```

Installation

Dependencies

tensorflow_cpp is a wrapper around the official TensorFlow C++ API. The C++ API including libtensorflow_cc.so must be installed on the system.

Instead of having to build the C++ API from source yourself, we recommend to check out our repository libtensorflow_cc. There, we provide a pre-built library and a Docker image for easy installation and usage of the TensorFlow C++ API.

Installation is as easy as the following. Head over to libtensorflow_cc for more details. ARCH=$(dpkg --print-architecture) wget https://github.com/ika-rwth-aachen/libtensorflow_cc/releases/download/v2.9.2/libtensorflow-cc_2.9.2-gpu_${ARCH}.deb sudo dpkg -i libtensorflow-cc_2.9.2-gpu_${ARCH}.deb ldconfig

If you have already installed the C++ API another way, you can use the provided TensorFlowConfig.cmake to enable the find_package(TensorFlow REQUIRED) call in tensorflow_cpp's CMakeLists.txt.

CMake

1. Clone this repository.

   bash git clone https://github.com/ika-rwth-aachen/tensorflow_cpp.git cd tensorflow_cpp

2. Install tensorflow_cpp system-wide.

   ```bash

   tensorflow_cpp$

   mkdir -p build cd build cmake .. sudo make install ```

3. Use find_package() to locate and integrate tensorflow_cpp into your CMake project. See the CMake example project.

   ```cmake

   CMakeLists.txt

   findpackage(tensorflowcpp REQUIRED)

   ...

   addexecutable(foo ...) # / addlibrary(foo ...)

   ...

   targetlinklibraries(foo tensorflow_cpp) ```

ROS/ROS2

1. Clone this repository into your ROS/ROS2 workspace.

   bash git clone https://github.com/ika-rwth-aachen/tensorflow_cpp.git cd tensorflow_cpp

2. In order to include tensorflow_cpp in a ROS/ROS2 package, specify the dependency in its package.xml and use find_package() in your package's CMakeLists.txt.

   xml <!-- package.xml --> <depend>tensorflow_cpp</depend>

   ```cmake

   CMakeLists.txt

   ROS

   findpackage(catkin REQUIRED COMPONENTS tensorflowcpp )

   ROS2

   findpackage(tensorflowcpp REQUIRED) amenttargetdependencies( tensorflow_cpp) ```

Testing

In order to build and run the test cases defined in tests/, execute the following.

```bash

tensorflow_cpp$

mkdir -p build cd build cmake -DBUILD_TESTING=ON .. make ctest ```

Documentation

Click here to be taken to the full API documentation.

The documentation can be generated by running Doxygen.

```bash

tensorflow_cpp/doc$

doxygen ```

Acknowledgements

This work is accomplished within the projects 6GEM (FKZ 16KISK038) and UNICARagil (FKZ 16EMO0284K). We acknowledge the financial support for the projects by the Federal Ministry of Education and Research of Germany (BMBF).

Notice

This repository is not endorsed by or otherwise affiliated with TensorFlow or Google. TensorFlow, the TensorFlow logo and any related marks are trademarks of Google Inc. TensorFlow is released under the Apache License 2.0.