Pose and Head Orientation Invariant Face Detection Based on Optimised Aggregate Channel Feature

Abstract

Face detection plays a critical part in the identification of people in biometrics. Face detection with the pose or head orientation image is really a challenging problem with surveillance applications since most techniques used in face detection only show better results when the front-facing images exist. The size of the image dataset is more and takes more time for training and detecting. The goal of our proposed work is to establish an effective method with minimum training time for detecting human faces, which can deal with both front and pose or head orientation faces. To accomplish this goal, (ACF) Aggregate Channel feature face detector with optimized parameters was proposed. Our proposed ACF face detector is a learning-based algorithm implemented with Pointing’04 and validated with FEI datasets and wild dataset. We implemented our proposed ACF face detector in MATLAB 2019a and compared their performance with other existing algorithms. The performance parameters of our proposed method such as average precision and log average miss rate are calculated, and the values are 0.90 and 0.18, respectively.