davidishe / Face_type_recognition

Squeezenet Example

This example shows:

• How to create an InferenceSession
• Load SqueezeNet model
• Use an image as input Tensor
• Run the inference using the input
• Get output Tensor back
• Access raw data in the Tensor
• Match the results with the predefined ImageNet vector

How to run

1. Download model file squeezenetV1_8.onnx from examples/models and put it in the current folder.

2. Start an http server in this folder. You can install http-server via

   npm install http-server -g
   

   Then start an http server by running

   http-server .. -c-1 -p 3000
   

   This will start the local http server with disabled cache and listens on port 3000

3. Open up the browser and access this URL: http://localhost:3000/squeezenet/

4. Click on Run button to see results of the inference run.

Files in folder

• index.html

  The HTML file to render the UI in browser

• index.js

  The main .js file that holds all ONNX.js logic to load and execute the model.

• squeezenet-piano.jpg

  A sample image chosen from one of the 1000 categories. Could be replaced with any image of your choice.

• squeezenetV1_8.onnx

  The ONNX model file that contains the SqueezeNet model.