We can see visualizations of neural networks in papers and literatures, which are very exquisite. You may wonder whether they are drawn using drawing software or office software. These are too low-end and inefficient for AI researchers, who use a number of open-source projects for neural network visualization. While each of the major deep learning frameworks provides corresponding visualization templates, they are not generalizable.
Netron is a neural network, deep learning and machine learning models viewer/visualizer and analyzer for Windows, Mac and Linux platforms, as well as Python web server and web browser. It was developed by Lutz Roeder from Microsoft, built on the basis of Electron, mainly implemented using the JavaScript language.
Netron can generate descriptive visualization for the model's architecture. It supports a very wide range of neural network model frameworks and formats, including ONNX, TensorFlow Lite, Caffe, Keras, Darknet, PaddlePaddle, ncnn, MNN, Core ML, RKNN, MXNet, MindSpore Lite, TNN, Barracuda, Tengine, CNTK, TensorFlow.js, Caffe2, and UFF. In addition, it also provides experimental support for PyTorch, TensorFlow, TorchScript, OpenVINO, Torch, Vitis AI, kmodel, Arm NN, BigDL, Chainer, Deeplearning4j, MediaPipe, MegEngine, ML.NET and scikit-learn.
In a real-world project, we will encounter various network models that require us to quickly understand its network structure. If we simply look at the model file, it is difficult to visualize the architecture of the network in our mind. In this case, we can use Netron, which enables us to clearly see the input/output of each layer and the overall architecture of the network. Maybe you have already used some deep neural network visualization tools, but Netron is probably your favorite so far. By contrast, it is not only omnipotent, but also easy to use.
Supported Frameworks
- ONNX (.onnx, .pb, .pbtxt)
- Keras (.h5, .keras)
- CoreML (.mlmodel)
- Caffe2 (predict_net.pb, predict_net.pbtxt)
- MXNet (.model, -symbol.json)
- TensorFlow Lite (.tflite)
- Caffe (.caffemodel, .prototxt)
- PyTorch (.pth)
- Torch (.t7)
- CNTK (.model, .cntk)
- PaddlePaddle(__model__)
- Darknet (.cfg)
- scikit-learn (.pkl)
- TensorFlow.js (model.json, .pb)
- TensorFlow (.pb, .meta, .pbtxt)
Sample Models
Sample model files to download or open using the browser version:- ONNX: squeezenet [open]
- TensorFlow Lite: yamnet [open]
- TensorFlow: chessbot [open]
- Keras: mobilenet [open]
- TorchScript: traced_online_pred_layer [open]
- Core ML: exermote [open]
- Darknet: yolo [open]
Download URLs
| License | Version | Download | Size |
| Freeware | Latest |  (mir) |
n/a |