Building tensorflow graphs and running deep learning predictions from your favourite visual programming interface
Tensorflow_Toolkit is an open source software library written in C# that exposes tensorflow
to visual programming interfaces like Grasshopper
and Dynamo thorugh the .Net bindings from https://github.com/migueldeicaza
This is a work in progress and aims at creating a high-level api to enable mass partecipation and fast experimentations with deep learning.
The code relies on the BHoM library, which reflects teh assembly to the visual programming platforms.
Let’s get started with Grasshopper.
cd
to your favourite folder and clone the necessary repositories. It is important that they are in the same folder.
git clone https://github.com/BHoM/BHoM
git clone https://github.com/BHoM/BHoM_Engine
git clone https://github.com/BHoM/BHoM_Adapter
git clone https://github.com/BHoM/Rhinoceros_Toolkit
git clone https://github.com/BHoM/Grasshopper_Toolkit
Compile the libraries above. They are based on the C# framework 4.6.1, so any modern Visual Studio version should be ok.
Compile the Tensorflow
Compile the Grasshopper_Toolkit again.
Done, you can open Grasshopper and start building your graph
The status of the code allows you to build a graph for now, and run a foward pass through a tensorflow session.
All the c++ tensorflow ops are exposed.
Keras is a good example of a high-level api that uses tensorflow as a backend.
Miming some of its high level functionalities would be a good start. More is to come.