openFrameworks addon for Tensorflow C++
Forks
Forks pushed to more recently than gilbertfrancois/ofxGFTensorflow itself, plus any admin-confirmed fork with commits of its own — worth checking if you're hitting issues with the original.
README
ofxGFTensorflow
About
This addon allows you to run Tensorflow models in your own openFrameworks projects, inspired by the OpenCV DNN module. The goal is to have a simple interface for loading a model and feeding data to the network. The addon library should work in any project, not just in a openFramework project. See remarks below how to modify it.
There is another excellent project ofxMSATensorflow by Memo Akten. Please have a look there as well and see what addon suits you best.
Dependencies
- ofxOpenCV
- ofxCv
- openFrameworks v0.10.0 (untested on lower versions)
Installation
- Go to the release page and download a stable version of the source _or_ type
git clone https://github.com/gilbertfrancois/ofxGFTensorflow.gitto install this repo inside your${openframeworks_dir}/addonsfolder. - Download the precompiled Tensorflow C++ libraries for your operating system from the release page and unzip the archive with headers and libraries.
- [Optional] Copy the file
./libs/tensorflow/lib/{linux64 | osx}/libtensorflow_cc.soto/usr/local/libto make your programs work in runtime.
ofxAddons