Visualizing your Neural Network with Netron

Neural networks, and especially the deep ones, have achieved many state-of-the-art results over the past few years. Many scholars and practitioners have used them to create cool tools and new techniques, which are used in various real-world scenarios today.

Let's say that you've identified a new type of architecture that works really well. Now, you wish to communicate about this architecture. How do you do so? And how can you visualize your neural network architecture easily - or inspect it, if you will?

Netron is such a tool. Being a viewer for neural networks and machine learning models, it generates beautiful visualizations that you can use to clearly communicate the structure of your neural network. What's more, using the tool, you can explore your models in great detail. And best of all, it's a cross-platform tool - which also means Windows and Mac support - and works with a wide range of machine learning frameworks and model formats.

In this blog post, we'll take a look at Netron. First, we'll discuss what it is and what frameworks and model formats it supports. Then, we move on to an example with Keras: we show you how to generate a Netron-ready model output, and how to visualize and inspect it subsequently.

Let's take a look! :)

Introducing Netron

Let's now take a look at Netron. Created by Lutz Roeder - from now on cited as Roeder (2020) - is a cross-platform tool for visualizing deep learning models, specifically deep neural networks.

Or as they describe their tool: Netron is a viewer for neural network, deep learning and machine learning models (Roeder, 2020).

It can generate beautiful visualizations of your neural network and supports a wide range of frameworks and formats. A slice from such a visualization can be seen on the right, and was generated from a Keras model.

Let's now take a look at the frameworks and formats that are supported by Netron. Then, we'll show you how to install the tool - which is really easy, and given the fact that it's cross-platform, it's supported for Windows and Mac machines as well.

Then, we continue by providing an example for Keras.

What frameworks and formats does Netron support?

As you can see, Netron supports a wide range of frameworks - and offers experimental support for a wide range of others (Roeder, 2020) :)

| Framework | Supported? | File types | |