Attribute detection with Body Camera Analytics
In the Body Camera Analytics profile, IBM® Video Analytics runs several attribute detectors in response to each captured person.
Properties and attributes
The attribute detectors that are run depend on which parts of the person are detected.
hairline, glasses, gender, age, facial hair, torso color, head color
Hair (any color)
|Facial hair||Facial hair
No facial hair
Solid color with logo
|Head Colors||Color 1
Each attribute that is detected is assigned a confidence score, which is a measure of the probability that the person has the attribute. For example, if the attribute is Male: 0% means a low probability that the person is male; 100% means a high probability that the person is male. Torso color is assigned multiple scores based on a palette of 13 colors.
When the property is age, the attributes, Old, Middle Aged, and Young are measured in relation to 50 years old. The confidence score is based on the predicted age. When you click View details, the value that is displayed indicates the predicted age. The usual range is approximately 10 - 80 years old.
Ranking shows all the results in ranked order, unless you select advanced mode for filtering and limit the number of results by filtering. As a filter on the ranked result, you can select not to display results below a certain threshold.
For example, when you select Young results are sorted by age in ascending order, the youngest people appear in the list first. The youngest person who is detected is in the first position and the oldest person is in the last position. Similarly, when you select Old results are sorted by age in descending order, the oldest people appear in the list first. Middle-aged is defined as the absolute distance from 50 and the results are ordered by the difference between the predicted age and 50.
Retrieving and searching video