If you are using one of our self-cleaning cameras, simply plug the camera into the ethernet (with PoE). If using one of your own cameras, you will need to connect the Edge AI Device on a network that has access to your camera. You can then visit the Edge AI Device web portal via IP address and complete the setup process.

Depending on your application or project, you may only need to record video for a few days before you can start training the camera.

Select a wide variety of images -- night, day, cloudy, rainy, strong shadows, etc. The more variety in images, the better the model will train. You want to aim to extract between 3,000 - 5,000 for this first dataset

Using the web portal, you can annotate images easily. For Image Classification (such as in the example above), you should be able to annotate 3,000 images in just a few hours.

Connect the camera or Edge AI Device to our Canopy Cloud Training Service or use an On-Premise Training Server to train a model using your annotated images. A typical model takes approximately 8 hours to train.

Once model training is complete, the model can be deployed onto the camera (or Edge AI Device). You should now start seeing AI predictions on the live stream in real-time and can export data via the web portal.