Best Practices for AI – Rhombus Support

In this article, we will discuss recommended practices and initial expectations for Rhombus Artificial Intelligence.

What does our AI do for Rhombus?
Initial AI expectations
Properly setting up the camera for optimal AI accuracy

What does our AI do for Rhombus?

The current artificial intelligence we use for Rhombus cameras helps track humans, vehicles, faces, and license plates. Our AI can identify the camera's humans, faces, vehicles, license plates, and unusual behavior. The AI compares faces and assigns an alphanumeric value to license plates in our cloud infrastructure.

Initial AI expectations

Some of our AI is stored/performed locally on the cameras. This means that when first registering the Rhombus camera you may experience some false positives for human or vehicle movement. If you experience a large number of false movements please feel free to reach out to the Rhombus Support team here for better assistance with this issue.

Properly setting up the camera for optimal AI accuracy

Below is a quick start guide for how to properly set up the camera to track particular movements and recognition for certain features. Below is a list of common tips that can be applied to the initial setup of the camera along with the expectations for AI features.

Framing of the camera view

When mounting and setting up a Rhombus camera you should always try to eliminate all unmoved space from the camera view. The camera should be mounted parallel to the horizontal plane. Here is a mounting guide for Rhombus cameras So if a camera is framed on a wall, you would not want the wall to be visible or limit visibility as much as possible. Below is an example of framing of a camera view with too much of the ceiling in the picture along with a fix.

Below is the same camera now adjusted to eliminate the ceiling from the frame.

Placing Activity regions

Setting up activity regions in a camera frame view helps by providing AI alerts only if the movement passes the threshold of that region. Also, setting up 'don't trigger' activity regions for areas such as bushes or any other non-human or vehicle movement that occasionally moves by environment changes. This helps improve your AI by reducing the number of false movement alerts. To learn more about activity regions see the article here.

Human Movement

For human detection, the camera will need to be positioned in a way where most of a human body shows in the camera's frame. If you only have a single body part in the frame initially the camera will only detect an 'Any Movement'. In order for the AI to detect human or vehicle movement, a certain number of pixels in the frame have to be changed.

Screen_Recording_2022-07-19_at_10.56.22_AM.gif

Facial Recognition

For facial recognition, we have a great guide here on how to manage facial recognition in both the aspect of positioning the camera and setting up the saved faces feature. For our AI to activate the facial recognition aspect it first needs to detect a human movement, then the AI will look for a face. Then the face will be sent to the cloud to compare it to saved faces.

Tip: We at Rhombus recommend that uploading saved faces for individuals is key to helping the AI compare the correct faces. See how to do that here. We also recommend that you look at our Objects and Distances page to better decide on what camera will accurately provide you with facial recognition.

Vehicle Movement

For vehicle movement, you will need to have the camera positioned so that the full car enters the field of view. If only portions of the car enter the field of view this may not trigger a vehicle movement detection from the AI. In order for the AI to detect human or vehicle movement, a certain number of pixels in the frame have to be changed.

License Plate Recognition

For license plate detection, we have a guide here on how to better position the camera for license plate recognition and how to add view and track license plates. For our AI to activate the license plate recognition feature it first needs to detect a vehicle movement, then the AI will look for a license plate on the car.

Tip: We at Rhombus recommend that if you want to have accurate LPR that you look at our Objects and Distances page to better decide on what camera will accurately provide you with license plate recognition.

Unusual Behavior

Our unusual behavior feature is a great feature to use to track people falling over, running through a building, or any other non-normal human movement. To configure the unusual behavior, please check out this article here.