TensorSpace: A Neural Network 3D Visualization Framework.

Dec. 18, 2018, 3:37 p.m. By: Kirti Bakshi


TensorSpace: Present Tensor in Space is basically a neural network 3D visualization framework that has been built by TensorFlow.js, Three.js and Tween.js.

TensorSpace provides APIs that are Keras-like in order to build deep learning layers, load pre-trained models, as well as generate 3D visualization in the browser. From TensorSpace, it is intuitive to learn what the model structure actually is, how the model is trained as well as how the model based on the intermediate information finally predicts the results.

After preprocessing the model, TensorSpace supports to visualize pre-trained model from TensorFlow, Keras and TensorFlow.js.

Why Tensorspace and what was the motivation behind it?

TensorSpace, as mentioned before is a neural network 3D visualization framework that has been designed for not only showing the basic model structure, but also presenting the processes of internal feature abstractions, intermediate data manipulations as well as final inference generations.

By the application of TensorSpace API, it is more intuitive to visualize and understand any pre-trained models built by TensorFlow, Keras, TensorFlow.js, etc. Apart from this,TensorSpace as well introduces a way for front-end developers to be rightly involved in the deep learning ecosystem.

Now, let’s take a look at the key features of Tensorspace:

*Interactive: Makes the use of Keras-like API in order to build an interactive model in browsers.

*Intuitive: Visualize the information from intermediate inferences.

*Integrative: Support pre-trained models from TensorFlow, Keras, TensorFlow.js.

When it comes to the working, The TensorSpace.js works well on Chrome, Safari as well as Firefox. TensorSpace is also compatible with mobile browsers. TensorSpace.org, the official site of Tensorspace provides documents, downloads and live examples of TensorSpace.js.

One can also find the source code of TensorSpace.js that is made to be available at Github.


  • Apache License 2.0

Now let us take a look at Applications of Tensorspace:

  • TensorSpace Playground: Online Deep Learning models that have been built with TensorSpace by tensorspace-team. All models in the "Playground" are interactive. One can move the mouse to see the relation lines among layers; can click the layer aggregation to check feature maps; and as well as move the camera to view from any direction

  • TensorSpace Hello World: CodePen TensorSpace MNIST-handwritten example by syt123450.

As an open source library, the team of TensorSpace, not limiting itself anywhere, also welcomes any further development on visualization applications.

For more information regarding the installation, documentation and more, one can refer to the links mentioned below:

Source and Information: GitHub

Official Link: Tensorspace