This repository contains several face detection algorithms. I'm using wrapper for all models to make those easier to implement.
class FaceDetectorWrapper:
def __init__()
def detect()
def detect_with_evaluation
The Pros and Cons of those algorithms are here
Example: Ground Truth: 8 faces
Models | FPS | Result |
---|---|---|
HaarCascade | 0.75 | 11 faces |
HoG Based Dlib | 2.17 | 8 faces |
CNN Based Dlib | 0.11 | 9 faces |