1

Is there a way to detect a face using opencv without using viola-jones algorithm?

Erick Campos
  • 13
  • 4
  • 1
    Is there a reason why you cannot use the viola-jones algorithm? – Quietust Apr 03 '17 at 19:17
  • I'm implementing a head pose estimation to college work. The algorithm starts with face detection (viola-jones) an after that it to trying to find the eyes, nose and track them through optical flow then finally occurs the head pose estimation. I've tested and it works, but sometimes the face detection fails due to the variance of illumination or somenthing like that. I just want to improve the face detection. – Erick Campos Apr 03 '17 at 20:51
  • I'm following this method: https://www.researchgate.net/publication/262320501_Real-Time_Head_Pose_Estimation_for_Mobile_Devices – Erick Campos Apr 03 '17 at 20:56

1 Answers1

1

An alternative to detecting faces using Haar Cascades in OpenCV is to use a Histogram of Oriented Gradients feature detector. HOG-based object detectors can be used for detecting faces, and this is the technique used in some other image processing libraries such as Dlib. A Histogram of Oriented Gradients feature detector stores a 2D grid of local gradients and gradient directions of an image and compares this with a reference grid trained on a data set.

samgak
  • 23,944
  • 4
  • 60
  • 82