tensorspace

Why we included this project

Explaining what a convolutional filter or dense layer really does is much easier when you can show someone a moving 3D model of the network. TensorSpace does exactly that: it renders pre-trained TensorFlow, Keras, and TensorFlow.js models as interactive browser scenes, letting you rotate around the layers, watch activations flow through, and inspect intermediate outputs instead of reading a static summary. It ships Keras-like APIs for building layers and a converter that prepares existing models, so you can load a network you already trained rather than constructing a toy from scratch. The payoff is clearest in teaching and internal demos, where a clickable, animated model conveys architecture and inference faster than any diagram. Just budget some time for preprocessing; the converter needs the model in the right format before the visualization comes together.