Custom Detection Events

The VisionAI Observ enables custom detection events, allowing you to set up any AI detection targets and logic. Unleash your imagination and break free from any application scenario limitations.

Quick Start

Customize detection events for the ultimate space and flexibility in AI. Click on the video below for a quick overview:

Creating Your Exclusive Detection Events

Enter "Detection," switch to "Event," view the current list of detection events in the scenario, and click "Edit" to edit an event.

Click "+ Event" in the upper right to enter the event editor, filling in the following items:

  • Event Name: Name of the event.

  • Event Description: Description of the event.

  • Scenario: The scenario to which the event belongs.

  • Model Repository: The AI detection model repository needed for the event.

Tip: The Model Repository relates to the AI model required for the event, dictating the AI detection object target. For optimal performance, currently, a single event can choose up to 2 model repository models for joint detection.

Click "Create" to add a new detectable event.

Finally, to actually detect custom events in a task, remember to include it in your set detection scenario.

Starting from Scratch? 🤔

If you're clueless about using the event editor, select "Example" in the upper right corner. It will provide various detection examples, which you can use to modify and create your own exclusive event detection logic.

Event Editor

The Observ event editor has powerful organizational features, allowing you to set up various detection logic and unleash your imagination, making AI no longer limited to any application scenario.

A detection event is composed of multiple event components. To determine whether a detection result triggers an event, it needs at least three components: "Object" + "Logic" + "Trigger".

It can be more complex combinations to determine whether a detection frame at a point in time is a triggering event. In Observ, we offer various logics, such as whether an object exists, whether there is overlapping repetition, count judgment, etc.


We can use "people" + "with" + "trigger" to create a detection event. This means when the AI detects people in the frame at a point in time, it will trigger the event.

The various event components in this example can be replaced or combined, for example, replacing "people" with "car" or "cat". Thus, whatever the AI detects, or whatever is needed for scene detection, you can define your needs in this way.

Due to the variety of combinations, we will introduce various event components in the following chapters.

Last updated