Skip to main content

Customize for Your Site

What You'll Learn in This Guide

After completing the Quick Start setup, you’ll now learn how to reduce false positives and increase detection accuracy in real-world operation.

  • Adjust per-object confidence thresholds

  • Refine detection scenarios for greater precision

  • Provide false-positive feedback to train EVA

1

Adjust Object Confidence Thresholds

Step Description

The confidence threshold determines how certain EVA must be before considering an object “detected.” Fine-tuning this value lets you optimize sensitivity for your specific site.

  • Increase the threshold (e.g., 0.5 → 0.8)
    • Makes detection stricter → fewer false positives
    • Recommended when you’re getting too many false alarms
  • Decrease the threshold (e.g., 0.7 → 0.4)
    • Makes detection more sensitive → catches more events
    • Recommended when important events are being missed
Threshold Adjustment Page
Threshold Adjustment Page

Actions

  1. Open the camera monitoring page
    • Select the camera you want to adjust
  2. Go to Threshold settings
    • In the Detection section, click “Details & Settings” next to the Scenario
  3. Select the target object
    • Click the object you want to adjust (e.g., “person”)
  4. Move the Detection Sensitivity slider
    • Slide right to reduce false positives
      Slide left to reduce missed detections
  5. Save changes
    • Click “Submit” to apply the new threshold
2

Refine Detection Scenarios

Step Description

Detection scenarios consist of Detection Steps (what to alert on) and Exceptions (what to ignore). Making these descriptions more specific dramatically improves accuracy.

  • Detection Steps
    • Define the exact conditions that should trigger an alert
  • Exceptions
    • Define normal situations that should not trigger alerts even if the main condition is met
Scenario Editing Page
Scenario Editing Page

Actions

  1. Click Edit Camera
    • Click the pencil icon on the camera detail page
  2. Navigate to Detection Scenario
    • Click “Next” in the Configuration section until you reach Detection Scenario
  3. Refine Detection Steps & Exceptions
    • Review the current scenario
    • Make Detection Steps more precise
    • Add clear Exceptions for situations that should be ignored
  4. Save changes
    • Click the “Save” button at the bottom of the page
3

Submit False Positive Feedback

Step Description

When EVA makes a mistake, tell it! Each false-positive report teaches EVA to automatically filter similar cases in the future — this is how EVA gets smarter over time.

  • How feedback learning works
    • You mark a detection as a false positive
    • EVA learns visual patterns from that image (lighting, position, shadows, etc.)
    • Future similar images are automatically suppressed
Feedback Submission Page
Feedback Submission Page

Actions

  1. Spot a false alert in the Commands panel
    • Alerts appear in real time with a thumbnail image
  2. Click “Feedback” on the incorrect alert
    • Click the “Feedback” button below the alert
  3. Review EVA’s reasoning
    • A popup shows EVA Detection Summary and its final decision
  4. Mark as false positive and explain why
    • Select “Yes, there is incorrect detection”
    • Write a short note, e.g., “The person is actually wearing a hard hat — angle caused the mistake”
  5. Submit the feedback
    • Click “Submit” — EVA learns instantly
Quick Start ExperienceNext: Advanced Configuration