Prototyping & Hardening for Production

Navigator is a versatile, visual, editor that allows you to easily create, run, and deploy AI Flows. With its lean interface, users can express complex logical flows with it that allow for a high degree of customization.

This article is a guideline to help you get started building your first, simple flows as well as exploring more advanced Flow setups.

Run a first Flow

A great way to start playing around with Navigator is to build a simple Flow that uses your Computer’s camera Input and sends it directly to an output that you can view on your screen.

Once you log in, click on the Element Drawer at the bottom left of the Canvas and search for the Camera Element. The Camera Element is a placeholder for any video Input. Navigator assigns provisions and assigns your default Computer Camera on Startup, so you should already see a “My Webcam” label on the Element. This tells you that the Camera Element will use your Computer’s camera as an input and provide it as its output for the next Element in your Flow.

Go back to the Element Drawer and add “Output Preview” to Canvas and place it to the right side of the Camera Element. Output Preview will simply display the input that is handed to it. Click and drag the connector dot on the right side of the Camera element to the connector dot on the left side of the Output Preview.

Now run the flow by clicking the Play button in the top right corner of the Canvas. Navigator will first install required dependencies on your machine*, if needed, and then run the Flow. Once the red Pause button appears instead of the spinner you know that the Flow is running. At the same time your built-in Camera should activate and the Output Preview Element will show an “Open” button. Click the Open button in the Output Preview Element to open a preview window that displays the video feed from the Camera Element.

To end the Flow, simply click the Pause Button in the upper right corner of the Canvas. Congratulations, you ran your first Flow!

Note: If dependencies need to be installed, it may take a moment for the model to start. The following model runs will start faster.

Run a first Inference

The previous example illustrates how to connect Elements and run a Flow in Navigator. But the real power of Navigator and the underlying Runtime is to execute AI Flows on your Hardware. In the following, we will build a simple Inference Flow that will detect Objects within the frame of your Camera.

Start with the Flow that you built in the previous step. We will keep the setup the same and add Inference in between. Open the Element Drawer, search for “Object Detector”, and drag it onto Canvas. This Inference Element runs YoLo under the hood to detect objects on frames. It takes frames as input and produces annotated frames (bounding box, probability, object label) as output. Rewire the Camera Element to connect to the Object Detector, and connect the Object Detector to the Output Preview input.

Make sure that the Camera Element has your Camera assigned and the other two Elements have your Computer assigned.

You can now run the Flow. Once the Open button shows up on the Output Preview, you can open it and watch YoLo detect objects in real time.

‍