Object Detection
Object detection is configured through three settings:
Common Object Type
A common object type is a specific type of object that EyesOnIt can detect. EyesOnIt V3 supports people, vehicles, animals and bags as common object types. Specifying a common object type is optional, but if you are looking for one of these types of objects, providing the common object type can improve accuracy.
Object Descriptions
EyesOnIt supports natural language object descriptions. If you provide a common object type, the natural language description refines the objects detected through the common object type detection. You could combine natural language object descriptions with common object types to look for a person with a red shirt, or to look for a blue truck.
Object descriptions include both “foreground” and “background” descriptions. You can provide one or more foreground descriptions that describe what you want to detect. You can optionally provide a background description for greater accuracy. If your background is easily described (in some cases the background changes often so it is not easily described), providing a background description helps EyesOnIt distinguish the object you care about from the background.
For processing video, foreground descriptions have related settings:
Alert
A true / false flag indicating whether EyesOnIt should generate alerts for this object description
Confidence Threshold
The confidence level threshold for alerts. Alerts will be sent if the confidence level exceeds this threshold.
Object Size
The object size tells EyesOnIt roughly how large an object is in your image or video frame. This helps EyesOnIt remove false positive detections. The object size is just an estimate. EyesOnIt will consider objects that are from 25% to 200% of the specified object size to see if they match your object descriptions. To determine the object size, capture a single frame from your video, load it into an image editor and measure its width and height with a rectangle selection tool. Use the larger value of width and height as the object size.