high performance computing on graphics processing units: hgpu.org

Convolutional Neural Networks (CNNs) are shown to perform very well in the areas such as video surveillance, object classification and face classification. Face classification has become pertinent to numerous applications, especially in this big data era of social platforms and social media. With the usage of unmanned air-borne vehicles like drones, the problem of face classification becomes very challenging because of the large visual variations. In this paper we introduce D-face, a novel application for face classification, and the input to the classification model is taken from the drone video sequences. In recent times, the CNNs have gained a significant importance in the class of classification problems, but the cons being its computational intensity. High performance computing using Graphical Processing Units (GPUs) has become an important tool in solving the compute intensive problems. We have used GPUs to speedup the classification rate for our problem. We bring about a faster parallel implementation that achieves almost 13x on GPUs compared to its serial implementation.