Haar Cascade to YOLOv8 Face Detection: 4.2x Speed Gain

#yolov8 #facedetection #opencv #haarcascade

The 2007 Algorithm That Quietly Runs in Production

Haar Cascades still power face detection in more production systems than you'd expect. OpenCV ships them by default, tutorials lean on them because setup takes three lines, and they run on a Raspberry Pi without breaking a sweat. But at some point, you hit the wall: false positives on tilted faces, missed detections in poor lighting, or inference that chokes when you scale from webcam demos to batch processing thousands of images.

I migrated a production face detection pipeline from OpenCV's haarcascade_frontalface_default.xml to YOLOv8n-face. The result: 4.2x faster inference on CPU, 89% mAP50 vs 67% on our validation set, and a 60% drop in false positives. This isn't a "deep learning is magic" post — YOLOv8 brought its own headaches, from preprocessing quirks to model export pitfalls I wish someone had documented.

Here's what actually changed, what broke, and when you should (or shouldn't) make the jump.