Skip to main content

Alerts

EyesOnIt tells you that it has detected something in your video feed that you care about by providing an alert. Alerts are only available when processing video. EyesOnIt can send these types of alerts:

  • a REST call to a Video Management System like Genetec or to other software that has a REST API
  • an SMS message (text only) to a cell phone number
  • an MMS message (text with an image) to a cell phone number
  • a visual notification within the EyesOnIt web UI

EyesOnIt alerts are based on the confidence levels of your object descriptions. To configure an alert, you tell EyesOnIt what confidence level is required to trigger the alert and for how long that confidence level must be observed. The alert can be reset when the confidence level for the alerting object description is below the threshold for a different period of time. Let’s look at an example. It is important to set the alerting options carefully so that EyesOnIt can send valid alerts when an object is correctly detected, and so EyesOnIt can avoid sending false alerts.

Continuing with our theme of detecting vehicles in a warehouse driveway, the image below shows the options that we care about from the EyesOnIt Video Monitoring and Alerting page in the web UI:

EyesOnIt alerts user interface

From this image of the video processing options screen, note the following:

  • there is an Alert Threshold next to the object description. In this example, the confidence that EyesOnIt has detected a vehicle must be 95% or greater before EyesOnIt generates an alert.
  • there is no threshold for the background description. EyesOnIt does not generate alerts for the background descriptions.
  • the Seconds Above Threshold Until Alert Fires field indicates how long the confidence level for an object description must exceed the Alert Threshold for that object description before EyesOnIt sends the alert. Based on the example above, EyesOnIt will send an alert after the confidence level that a vehicle has been detected exceeds 95% for five seconds.
  • the Seconds Below Threshold Until Alert Resets field indicates how long the confidence level for an object description must be below the Alert Threshold for that object description before EyesOnIt can consider sending another alert. Based on the example above, once an alert has been sent for a detected vehicle, EyesOnIt will not consider sending another alert for a vehicle until the confidence level that a vehicle has been detected is below 95% for 10 seconds.

These alert settings are critical to help EyesOnIt generate accurate, reliable alerts. If the alert threshold is too low, EyesOnIt may send an alert when an object was not actually present. If the alert threshold is too high, EyesOnIt may not send an alert when an object was actually present. The time settings help ensure that EyesOnIt sends one alert and only one alert for each event. Some trial and error may be necessary to find the right values for these fields.

Tip: to determine the right confidence levels for alert thresholds, capture two frames from your video. The first frame should be captured when the object to detect is present. The second frame should be captured when the object to detect is not present. Save these frames as images and then open those images one at a time in the EyesOnIt Object Detection section of the EyesOnIt web UI. You can experiment within this UI to find the detection regions, object size and object descriptions that produce the best results. The best results are where EyesOnIt consistently gives a confidence level above a certain threshold when the object to detect is present, and a confidence level below that same threshold when the object is not present. Try to find settings where the gap between these confidence levels is as high as possible. You can verify your confidence levels by saving more frames from your video with and without the object. Make sure to include frames with all of the lighting conditions that you expect to encounter in the real world. When you find settings that produce good confidence levels under all conditions, you can be confident that these settings will work well to process your video.

Also based on the image above, these additional fields are important:

  • the Phone Number for SMS Alerts field allows a user to enter a phone number to receive SMS alerts (with or without an image). This phone number needs to start with a country code like +1 for the US. This field can be blank if the user does not want SMS alerts.
  • the Include Image with Alert field allows the user to specify whether they want to receive an image with their SMS alert. If this field is checked, a valid phone number must be provided.