There are 3 key things to remember when creating a dashboard

1. Select chart types that best visualize the type of Event Detections you’ve generated 

2. When visualizing data from multiple video sources, it’s important to select related data types
   

3. When applying filters, you can either filter data displayed in individual charts, or apply filters across all charts within your dashboard
   

Adding charts to your dashboard

After creating a dashboard you will be taken to the dashboard homepage. The next step is to start populating the dashboard with charts. Let’s walk through how this is done.

Navigate to the ‘Charts’ sub-heading and click the ‘+’ button next to it (top left corner) to begin adding charts to your dashboard.

There are 3 main steps in creating a chart for your dashboard
(see the three steps outlined at the top of the pop-up box in within the below screenshot): 

1. Select a chart type that best suits the video events you want to analyze

2. Select the processed footage that you want to visualize

3. Configure your charts with formatting, data grouping and filters

Types of charts

Now, let’s explore the different chart categories available in VisualCortex and when they’re most suited.  

Heatmaps are used to visualize and communicate density patterns over time (pedestrian or vehicles) for a fixed, single field-of-view.

Charts under the sub-heading ‘Directional Count’ are used to visualize the number and directional flow of selected objects that cross a defined virtual lines. Count data can be segmented by object class. Counts can also be aggregated from across multiple video sources.

Charts under the subheading ‘Dwell’ can be used to understand the amount of time that objects (such as pedestrians or vehicles) spend in a defined zone. Once more, count data can be segmented by object class, with counts also able to be aggregated from across multiple video sources.

Chart set-up: Heatmaps

In this section, we showcase the steps required to create a Heatmap chart for dashboards. 

To begin, navigate your mouse to the Heatmap icon and click the left mouse button. 

After selecting Heatmap as your chart type, a pop-up window will appear. That window contains a folder structure in which your processed video footage is contained: This is step two of the set-up process, where you select your Data Sources. 

Begin by selecting the Virtual Camera, and / or Event Definitions associated with that footage, from which you’d like to generate a heatmap. Then, navigate to the ‘next’ button on the bottom right of the pop-up.

The third step is to configure the heatmap chart. 

Within the configuration options, give the new chart a practical name that describes the type of footage and detections being visualized. The name of the chart will appear within the dashboard. The name can be changed later too.

Next, select a grid size for your chart. These dimensions can be additionally adjusted once the chart is added to the dashboard.

Multiple Machine Learning Models can be run on the same camera. Select the Model type and size that you would like to display Event Detections for within your chart. Large Models render objects / actions at a more granular level of detail, but cost more credits to initially process.

Lastly, you can operationally choose to filter detections by assigned object class (eg: Only count the number of buses, excluding all other vehicle types).

After clicking the blue ‘Save’ button in the bottom right corner of the pop-up, the chart should appear on the dashboard.

Chart set-up: Directional Counts

Navigate to the ‘Charts’ sub-heading and click the ‘+’ button next to it (top left corner) to add another chart to your dashboard.

Within the ‘Select Charts’ pop-up, navigate to the ‘Directional Count’ category of charts. In this example, we will demonstrate how to create a line chart to visualize directional count data. 

After selecting ‘Line’ as your chart type, a second pop-up window will appear. As with the heatmap set-up example preceding this, the second ‘Data Sources’ window presents a folder structure in which your processed video footage is contained. Select the Virtual Cameras, and the ‘directional count’ Event Definitions within them, from which which you’d like to generate a your chart. Then, navigate to the ‘next’ button on the bottom right of the pop-up.

As with the heatmap example, you can name your chart, select the Model you want to apply, and the object classes you want to visualize. 

For Directional Counts, you can also group the way in which you visualize data in two ways: 

1. By Class

2. By Direction 

Note: You can modify your selected data grouping later if desired.

Once your desired configurations are selected, press the blue ‘Save’ button, located in the bottom right of the pop-up.

A note on interpreting Direction Line charts:

• The x axis represents the time period 

• The y axis represents the unique count of objects

Chart set-up: Combining multiple video sources to visualize aggregated metrics

Next, let’s explore how to create aggregated metrics by combining detections from multiple video sources. Again, navigate to the ‘Charts’ sub-heading and click the ‘+’ button next to it (top left corner) to add another chart to your dashboard.

In this demonstration, we’ll use a table to display aggregated metrics. To do this, navigate to the ‘Directional Counts’ category of chart types, and select ‘Table’.

When selecting your ‘Data Sources’, be sure to choose related footage to generate meaningful aggregations (such as total counts across multiple highway entries or exits).

As with the heatmap and line chart examples, you can name your chart, select the Model you want to apply, and the object classes you want to visualize. 

The key difference is the ability to ‘Select pivot options’ – the way in which you will allow users to view and toggle your grouped data.

Once your configurations are finalized, press the ‘Save’ button in the bottom right corner to produce a table chart based on your selections.

Below is an outline of how to interpret data from the table chart produced: 

• Directive is the source of the data

• Time period is the time the data was recorded

• Class tabs (‘bus’, ‘car’, ‘heavy_goods_vehicle’) show the number of entries and exits recorded for processed footage from each Virtual Camera, sorted by detection type

Chart set-up: Dwell charts (Histogram)

Here, we step through creating a Histogram Chart, in order to produce a heatmap that can visualize dwell-time. This chart type displays the duration that specific objects spend within a defined zone over a selected timeframe.

Navigate to the ‘Charts’ sub-heading and click the ‘+’ button next to it (top left corner) to add another chart to your dashboard.

Within the ‘Select Charts’ pop-up, navigate to the ‘Dwell’ category of charts. Select ‘Histogram Heatmap Chart’.

As with the chart type example, you can name your chart, select the Model you want to apply, and the object classes you want to visualize. 

For Histograms, you also have the option to select the ‘Duration’ – the intervals over which you want data visualized within the chart. Your chosen duration should align with your footage length and the volume of Event Detections you want to visualize and communicate over that time period. 

For example, if the number of detections is high, or the length of the video short, aggregating by smaller intervals will generally make the most sense.

A note on interpreting Histogram charts:

• The X axis represents the time period in which the object was detected

• The Y axis represents the duration the selected class of object has spent within the dwell zone

Once you’ve added the individual charts you desire and have built your dashboard, you can dynamically adjust the time period – and the granularity by which data is displayed – via the filters in the top right corner.

To resize a dashboard item, left click and hold the bottom right arrows of the object as you drag your cursor across the screen.