Exercise - Bring your AI model to deploy a video analytics solution

Completed

Now that you have deployed the YOLO model to the edge device, you can deploy the Vision on Edge solution model.

Make sure that you have the followings:

  • Prediction endpoint that has form http://{module-name}:80/score
  • tag.txt file downloaded that has object labels
  • RTSP URL that has form rtsp://rtspsim:554/media/<video-name>

Connect to web application

In this exercise, you'll connect to the Vision on Edge solution web application. Firstly, you'll add a camera that streams out the video.

  1. Open your browser, connect to http://<replace_your_public_ip_address>:8181.

  2. To find your Public IP address, go to the overview page of your virtual machine and copy it.

    The illustration shows overview page of virtual machine.

  3. When you connect to Vision on Edge solution, select the Home page.

    The illustration shows edge solution.

Add a camera

  1. In this step, you'll add a camera that streams out the video feed.

  2. Navigate to the Cameras page and add a camera.

  3. Fill in the required information:

    • Camera name: Give a name to your camera
    • RTSP URL: Enter URL that has form of rtsp://rtspsim:554/media/<video-name>
    • Location: Create a location or choose an existing one

    The illustration shows how to add a camera.

Upload your pre-trained YOLO model to the Vision on Edge solution

  1. To bring the AI model to the platform, select on the Models and select on '+' to start.

  2. Fill in the required information:

    • Module name: Give a name to your model
    • Endpoint: Enter prediction endpoint that has form of http://{module name}:80/score
    • Labels: Upload tag.txt file
    • Header: Leave it empty if not necessary
  3. Select Add.

    The illustration shows how to add a machine learning model.

Create a solution deployment

  1. Navigate the Deployment page.

  2. To deploy a task, fill in the required information:

    • Give a name to your deployment
    • Choose Model
    • Select Camera
    • Select Objects
  3. Select Deploy.

  4. You'll see that the gaps will be identified in the video stream.

    The illustration shows how to deploy a task.

Check your knowledge

1.

If you are a customer and you want to use one platform that will allow you to deploy different machine learning models for different scenarios, Vision on Edge is a good solution template. Because: