Mini Project 1 · Danger Zone Monitoring

1. Overview

In the Operation chapter we read video frame by frame with a While Loop and branched a Flow with If-Then. This Mini Project puts both to work on something real: watching a danger area on a work site, and raising an alert the moment somebody walks into it.

What you'll build is a Flow that reads a video, detects people in every frame, then checks whether any of them are standing inside a zone we draw on the picture. If someone is, the Dashboard reports how many and switches the status to say a person was found in the danger area. If the area is clear, the count goes back to 0 and the status returns to safe.

The Node that makes this work is Detected Zones Counters. It takes the boxes from a detection Node and tells you what is sitting inside each zone.

2. Learning Objectives

By the end of this project, the learner will be able to:

• Read video with the Server Media Input Node and run detection frame by frame
• Use General Object Detection with NARA's pre-trained model to detect people
• Configure the Detected Zones Counters Node and draw zones on a reference image
• Read countInside and hasObjects and use them to build conditions
• Branch a Flow with If-Then so the status of the area follows what is in the zone
• Lay out a Dashboard showing video, count, and status together

3. Prerequisites

• Completed the basic course and able to build a Flow from scratch (see Build Your First Flow)
• Completed the Operation, since this project uses both the While Loop and If-Then
• Comfortable with General Object Detection and the Variable Nodes
• A video file or image to test with. This example uses footage of workers pouring concrete, but any clip that suits your own use case works

4. Expected Result

After deploying and pressing Start on the Dashboard, the video plays with detection boxes on it and the zone drawn over the top. As soon as someone walks into the zone, the number goes up and the status changes.

5. System Architecture

Everything happens inside NARA, with nothing external to connect to. The video is the input, detection and zone counting are the processing, and the Dashboard is the output.

The thing to watch out for is that a single pass reads only one frame, so the whole chain has to sit inside a While Loop.

6. Flow Architecture & Nodes

The Flow uses the Nodes below. Anything that is basic configuration or left at its default is not spelled out again.

On the Dashboard side there are 4 Widgets:

• Basic Display × 2, for the Danger Zone status and the number of people in the zone
• Trigger Button × 1, for Start
• Image Frame × 1, for the video

7. Key Concepts

Detected Zones Counters checks and counts objects inside an area you define on the picture, such as a danger area, a work area, or an inspection point.

The important thing is that this Node does not detect anything itself. It takes bounding boxes from a Node such as Object Detection or Object Tracker, compares each one against the zones you drew, and reports the result zone by zone.

There are really only three things to set:

And the values the Flow reads back out:

• frameOut image: the picture after processing, with the zone, reference point, and count drawn on it. This is what goes to the Dashboard
• hasObjects boolean: whether this frame has at least one object in the zone
• countInside number: how many objects are in the zone in this frame
• labels array: the labels of those objects, e.g. ["person", "person"]

For the full parameter list and output schema, see the Detection Count Zones Node reference.

8. Guided Workshop

Part 1 · Read the video and detect

Start with just the video reading and person detection. When you create the While Loop, remember it needs its End connected, otherwise the pass never completes.

Configure the Nodes one at a time:

Node 1 · While Loop

Nothing to change yet. Leave Loop Delay at 0 for now; you can adjust it later.

Node 2 · Server Media Input

Load the video or image file you want to run detection on.

Node 3 · General Object Detection

Deploy and check the Dashboard. The video should play, and people in the area should be detected.

Part 2 · Add Detected Zones Counters

Now add the Node that creates the area to watch. What we want to watch here is the people coming out of General Object Detection.

This is what the Node looks like before anything is configured:

Wire it in and the Flow ends up looking like this:

As mentioned earlier, this Node needs both Input Image and Rectangular Objects, and both can reference General Object Detection directly:

Part 3 · Draw the zone

Put the downloaded picture into the Zones section of Detected Zones Counters, in the field marked Drag and drop an image here or click to select a file. Then press + to create a zone. Once you finish drawing, you can pick which zone it should be. One picture can hold several zones.

Leave the other settings at their defaults for now. You can come back and change them later.

Deploy and look again. People inside the zone are picked up, and it tells you how many were counted. Here Zone 1 counts 2.

At this point the Detected Zones Counters Flow is working.

Part 4 · Make it do something with the result

Now to make the Flow a bit more capable: count the people in the area and show how many there are, and treat an occupied area as dangerous, worth raising an alert for. If nobody is there, the area counts as safe.

In plain terms: if there is a person in the Danger Area, the status flips to say so and the number of people detected is shown. If not, the status goes back to safe and the count is 0.

We'll use the Variable Nodes created earlier, with Count branching off to Reset Count and Counting. First, set the condition inside If-Then, comparing whether countInside of Zone 1 is greater than 0:

Then wire the branches as shown: countInside > 0 leaves through Then into Counting to record the number of people, and on to the Danger Zone Node set to true. What you do with that status next, whether that is a siren, a notification, or driving some other equipment, is up to you; this example only displays a message for now.

Part 5 · Configure the Variable Nodes

Count Node uses type Number.

Counting Node

Reset Count Node

Danger Zone Node uses type Boolean.

Two Nodes write into Danger Zone. They are configured identically apart from a single switch.

Set the second one up exactly the same way, but leave the Variable Value switch off so it writes false, and wire it to the Else path for when the zone is empty.

Part 6 · Finish the Flow

Wire the Flow together as shown, and remember it has to end at the Loop End, so every path has to lead back to the While Loop's End.

One more thing to go back and fix: Variable Modify Image has to reference Detected Zones Counters as its Source rather than the detection Node, otherwise the zone never shows up on the Dashboard.

Part 7 · Lay out the Dashboard

Now build the Dashboard side with the Widgets needed to show the results:

9. Flow Explanation

Once deployed, pressing the Trigger Button enters the While Loop, and one pass through the loop handles exactly one frame.

Server Media Input hands a frame to General Object Detection, which returns the bounding boxes of the people it found. Detected Zones Counters takes those boxes, compares them against Zone 1, and sends out countInside along with a frameOut picture that has the zone drawn on it.

If-Then then reads countInside. Above 0, the Then path writes that number into Count and sets Danger Zone to true. At 0, the Else path resets Count to 0 and sets Danger Zone to false. Either way the path has to end at the While Loop's End before the next pass can begin.

The reason Count and Danger Zone are created once but modified in two places is the same pattern as the Operation chapter: Create declares the variable, and each branch of the condition writes its own value into it.

10. Testing Scenarios

Normal case: Before anyone walks in, the count is 0 and the status reads as working. As people enter the zone, the count moves with however many are inside, and the status flips to stopped. Once they all leave, both values return to their starting state.

Additional test cases:

• Move the zone to a different part of the picture and see whether the count follows the new area
• Change the condition from countInside > 0 to countInside > 1, so one person still passes but two do not
• Draw a second zone and count two areas separately using zone2
• Change Label names to detect to something else in the footage and see how the count changes

11. Troubleshooting

• The zone sits in the wrong place on the picture: Check that the width and height in the Zones section match the resolution of the video, and that the reference image is the one downloaded from the Image Frame Widget.
• People walk through the zone but the count stays at 0: By default the middle of the bounding box is what decides. A box that only overlaps the edge is not counted until its centre moves inside. Check also that Rectangular Objects really does reference detectedObjects.
• Detection boxes show but the zone does not: Variable Modify Image is still pointing at the detection Node. Change its Source to frameOut from Detected Zones Counters.
• The video plays once and freezes: A Node inside the loop is not connected back to the While Loop's End, so the pass never completes.
• The status never changes: Check that the If-Then Data Type is Number, and that Then and Else lead to the right Variable Modify Nodes.

12. Challenges and Summary

Challenge 1: Add a second zone for a different area, such as a walkway, and show its count on the Dashboard alongside the first, using zone2.countInside as the source.

Challenge 2: Use labels instead of countInside so the Flow reacts differently depending on what is in the zone, so a person counts as dangerous while other objects are let through.

Challenge 3: Instead of only displaying a message, drive something outside NARA when someone is in the zone. Mini Project 2 covers watching a value and acting the moment its status changes, and the same condition works just as well for an MQTT message or a Modbus write.

Project Summary: This project connected detection to a decision. The video is read frame by frame, a model finds the objects, Detected Zones Counters narrows the result down to the area we care about, and If-Then turns that into a status the Dashboard can show.

The same structure works for anything defined by an area rather than by the whole picture: counting items at an inspection point, watching a loading bay, or checking whether a walkway is clear.