Visualizing an ONNX model using Netron
The ailia SDK, an inference framework for edge devices, uses ONNX to perform fast GPU-based inference. In this article, we will present our findings on the visualization of ONNX models obtained in the process of developing the ailia SDK.
About Netron
Netron is a cross-platform machine learning model visualization tool that allows you to visualize the structure of your model by simply uploading an ONNX file or Prototxt.
Usage
Click on Open Model and specify ONNX or Prototxt.
Once opened, the graph of the model is displayed. By clicking on the layer, you can see the kernel size of Convolution and the names of the INPUTS and OUTPUTS blobs.
Analyze a model from a URL
You can also open the model from the URL parameter.
As an example let open the BlazeFace
model using the URL below.
The README in the ailia MODELS repository provides Netron links for various models such as ResNet50 and YOLOv3, so please feel free to use them to analyze your model architecture.
ax Inc. has developed ailia SDK, which enables cross-platform, GPU-based rapid inference.
ax Inc. provides a wide range of services from consulting and model creation, to the development of AI-based applications and SDKs. Feel free to contact us for any inquiry.