AI Confidence Settings

In this article, we will cover the following:

• AI Confidence Thresholds
• Adjusting AI Confidence
  • Increasing Confidence
  • Decreasing Confidence
• Environmental Impacts to AI Performance
• Troubleshooting AI
• Helpful Links
• Contact Support or Sales

AI Confidence Thresholds

In AI detection models, confidence refers to the AI’s certainty about its identifications. Higher confidence indicates the AI has higher certainty that its identification is correct.

AI confidence thresholds are used to determine how certain an AI model is about its identifications. Setting appropriate thresholds can help balance the trade-off between false positives and false negatives, ensuring that the AI system behaves reliably and safely.

• High Confidence: The more strict the model is when identifying something as a human or a vehicle; risk of missing some valid targets.
• Low Confidence: The less strict the model is when identifying a human or vehicle; risk of incorrect identifications.

High Confidence Setting
When the system's confidence is set high, it means the model is very strict about what it considers to be a human or a vehicle. It will only make an identification if it is very sure that it’s correct. However, if the confidence level is set too high, the system might miss some valid detections because it demands a very strong confirmation before identifying something.

Example: Imagine you're using a security camera with human detection to identify people. If you set the confidence threshold at 95%, the system will only recognize someone as a person if it’s 95% sure that the movement in the clip matches the model for human movement. If the person's is partially obscured or if the image quality is poor, the system might not recognize them even though they are actually there. This can lead to missed detections of people who should be identified.

Low Confidence Setting
When the system's confidence is set low, it means the model is less strict and will identify things as humans or vehicles even if it’s not very sure. While this can help in detecting more potential targets, it also increases the chance of false positives, where the system might incorrectly identify something as a human or vehicle when it actually isn’t.

Example: Imagine you’re using the same security camera with human detection, but you set the confidence threshold at 50%. The system will now recognize movement as a person as long as it’s 50% sure about the identification. This means it might identify a random object, like moving tree branches or a bird, as a person because it’s less strict when comparing the movement to the model. As a result, you’ll get more alerts and notifications, but many of them might turn out to be false alarms.

Adjusting AI Confidence

Finding the correct threshold value for a given device can be challenging. While adjusting the confidence threshold can improve the performance of the AI model, external factors such as ambient lighting, camera positioning, and camera angling will significantly impact overall performance.

AI Confidence can be adjusted for the following event types:

• Human Detections
• Face Detections
• Vehicle Detections
• License Plate Detections

In order for the AI model to identify faces, a human must first be identified in the frame. Similarly, a vehicle must first be identified in the frame to identify license plates.

Increasing Confidence

If one increases the confidence threshold, the AI will scrutinize the data more in order to reduce false positives. However, going too far in this direction can lead to the AI filtering out event detections.

Decreasing Confidence

If one decreases the confidence threshold, the AI will be more lax with its identification criteria to reduce the number of missed events. This increases the number of detections being made, however, going too far in this direction can lead to false positives.

Environmental Impacts to AI Performance

Successful AI detections rely on an appropriate environmental setup. All camera setups are susceptible to environmental factors that can reduce the effectiveness of the AI model. Please note that each detection type may require a different environmental configuration.

One key environmental factor is external lighting. Adequate lighting is vital for cameras to identify objects such as people, faces, vehicles, or license plates. Without sufficient lighting, our cameras may not be able to make these identifications. Additionally, the height and angle at which a camera is mounted can also affect detection accuracy. Click here for more details. 

Troubleshooting AI

Please Note: AI detections are heavily dependent on camera positioning. Before troubleshooting AI, ensure your camera is mounted and angled appropriately for the event type you try to detect. For more details on optimal camera positioning for specific event types, click here.

Helpful Links

• Best Practices for AI
• Enable AI Bounding Boxes
• Features
• Optics and Object Distances for Analytics
• Managing Facial Recognition
• Managing License Plate Recognition (LPR)
• People and Vehicle Counting

Contact Support or Sales

Have more questions? Contact Rhombus Support at +1 (877) 746-6797 option 2 or support@rhombus.com.

Interested in learning more? Contact Rhombus Sales at +1 (877) 746-6797 option 1 or sales@rhombus.com.