netron

Netron is a viewer for neural network models.

Netron loads ONNX models (.onnx or .pb), Keras models (.keras, .h5 or .json) and TensorFlow Lite models (.tflite) and has experimental support for TensorFlow models (.pb and .meta).

Install

macOS

Download the .dmg file or with Homebrew run brew cask install netron

Linux

Download the .AppImage or .deb file. The .AppImage needs to be made executable after download.

Windows

Download the .exe file.

Download Models

Below are a few model files you can download and open:

ONNX Models

• Inception v1
• Inception v2
• ResNet-50
• ShuffleNet
• SqueezeNet
• VGG-19
• BVLC AlexNet

TensorFlow Lite Models

• Smart Reply 1.0
• Mobilenet 1.0 224 Float
• Inception v3 2015
• Inception v3 2016 Slim

TensorFlow models

• Inception v3
• Inception v4
• Inception 5h

Install Python Model Server

To run Netron in a web browser, install the Python web server using pip:

pip install netron

Launch the model server:

netron my_model.onnx

To serve a model from Python code:

import netron

netron.serve_file('my_model.onnx')