A face tracking framework based on convolutional neural networks and Kalman filter

This paper presents a method for real-time detection and tracking of the human face. The proposed method combines the Convolution Neural Network detection and the Kalman Filter tracking. Convolution Neural Network is used to detect face in video, which is more accurate than traditional detection method. When the face is largely deflected or severely occluded, Kalman Filter tracking is utilized to predict the face position. The objective is to increase the face detection rate, while meet the real time requirements. Our method is implemented based on Caffe framework. The experimental results show that our method achieves superior accuracy over the existing techniques and keeps real time performance.