Sensing and privacy
Computer vision
Vision par ordinateur
Computer vision is the automated interpretation of visual information from images or video. Curb systems may classify vehicles or estimate occupation, while lighting, occlusion, camera angle, training data, and privacy controls influence reliability; a classification output should be distinguished from a verified account of the observed activity.
Understanding the term
Test performance during night conditions and crowding, not only on clear demonstration images. Classification error can vary between vehicle types and can distort category-specific curb statistics.
Validate the measurement against an independent reference under representative conditions. Record uncertainty, coverage, missing intervals, and the difference between a raw observation and an inferred activity. Technical outputs should be used only for purposes they can support reliably. Assign responsibility for calibration, corrections, access, and retention, and assess privacy across the complete data flow. An operational dashboard can conceal measurement errors if it displays precise values without their method or limitations.