Haar Cascade Classifier

Updated 25 Sep 2026

•
What it is
- •
  a cascade of learned classifiers using Haar-like image features to detect objects efficiently across image locations and scales
•
Cascade principle
- •
  early stages reject many non-object windows cheaply; later stages examine the remaining candidates
•
Lab parameters
- •
  scaleFactor - step between image scales
- •
  minNeighbors - support required when grouping candidate detections
- •
  minSize - minimum detection size
•
Caution
- •
  these settings affect speed and false detections; minNeighbors is not a probability confidence score
- •
  a frontal-face model has limitations under pose, illumination, and occlusion changes
•
Source
- •
  https://docs.opencv.org/4.x/db/d28/tutorial_cascade_classifier.html
•
Related
- •
  Face Detection

Linked references 1