Common Detection Errors
Understanding common face detection errors helps you debug and improve your system.
6 min•By Priygop Team•Updated 2026
Types of Errors and Fixes
- False positive (ghost face): detected a face where none exists. Fix: increase minNeighbors, add face validation (check for eyes within the region)
- False negative (missed face): a real face was not detected. Fix: decrease scaleFactor (try 1.05), decrease minNeighbors, check image quality
- Too many overlapping boxes: multiple boxes for the same face. Fix: apply non-maximum suppression or use groupRectangles()
- Detection fails on side profiles: frontal cascade only detects forward-facing faces. Fix: also load haarcascade_profileface.xml and combine results
- Fails in low light: preprocessing helps. Fix: apply equalizeHist() before detection
- Very small faces not detected: Fix: decrease minSize parameter or upscale the image before detection
Diagram
Loading diagram…
Machine Learning follows a structured pipeline from data to deployment
Key Takeaways
- Understanding common face detection errors helps you debug and improve your system.
- False positive (ghost face): detected a face where none exists. Fix: increase minNeighbors, add face validation (check for eyes within the region)
- False negative (missed face): a real face was not detected. Fix: decrease scaleFactor (try 1.05), decrease minNeighbors, check image quality
- Too many overlapping boxes: multiple boxes for the same face. Fix: apply non-maximum suppression or use groupRectangles()