Fast GPU and CPU computing for Head Position Estimation

Abstract

Abstract-The head movement based control methods in the 3D graphic applications requires the real-time face position estimation. Therefore, the tracking method at the high speed and with the minimal latency is needed. This is especially hard to achieve when the face is tracked with the use of the high resolution video image on mobile devices. In the article, we present several methods for an acceleration of the face position estimation method based on the fuzzy skin color classifier and other color-based face tracking methods. The acceleration is achieved through a highly parallel GPU computation, the precalculation of the classifier weights and through the combined computations on the GPU and the CPU. The achieved computation time is independent of the used skin color classification method, allowing for use of very complex classifiers. The presented methods provides the robust head position tracking on the high resolution video image of 1920x1080 pixels, at 300 frames per second, on the mobile device with a low computing power.

Face Occlusion Detection Using Skin Color Ratio and LBP Features for Intelligent Video Surveillance Systems

“…To address the problem, GPU technique is used to accelerate process by using a parallel implementation of the HOG algorithm. Comparing CPU implementation, GPU based on parallel architecture has a better computational performance [17], [18]. The result of head detection demonstrates that our implementation using GPU can achieve a speedup of over 5 times, which allows for real-time detection.…”

A compact deep convolutional neural network architecture for video based age and gender estimation

“…It typically consists of a face detection [1] and the actual estimation module. The designed vision system must be resistant to changes in face appearance (facial expressions, hairstyle, presence of beard or moustache, glasses and to some extent also make-up), different lightening conditions, inaccurate face localization in the image, face orientation (frontal, from profile and also rotated), size of the input image and it's quality (presence of noise, blur caused by movement of the person, underexposure, overexposure, shadows, etc.…”