项目作者: epignatelli

项目描述 :
Building tensorflow graphs and running deep learning predictions from your favourite visual programming interface
高级语言: C#
项目地址: git://github.com/epignatelli/TensorflowVisual.git
创建时间: 2018-12-03T23:45:21Z
项目社区:https://github.com/epignatelli/TensorflowVisual

开源协议:GNU Lesser General Public License v3.0

下载


Hi, I am TensorflowVisual

License: LGPL v3

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.


To get started

The code relies on the BHoM library, which reflects teh assembly to the visual programming platforms.
Let’s get started with Grasshopper.

  1. cd to your favourite folder and clone the necessary repositories. It is important that they are in the same folder.

    1. git clone https://github.com/BHoM/BHoM
    2. git clone https://github.com/BHoM/BHoM_Engine
    3. git clone https://github.com/BHoM/BHoM_Adapter
    4. git clone https://github.com/BHoM/Rhinoceros_Toolkit
    5. git clone https://github.com/BHoM/Grasshopper_Toolkit
  2. Compile the libraries above. They are based on the C# framework 4.6.1, so any modern Visual Studio version should be ok.

  3. Compile the Tensorflow

  4. Compile the Grasshopper_Toolkit again.

  5. Done, you can open Grasshopper and start building your graph


What can you do right now?

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.


What is the road map?

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.