Search

Using machine learning in high-speed networks for tasks such as flow classification typically requires either very resource efficient classification approaches, large amounts of computational resources, or specialized hardware. Here we provide a sketch of the discretize-optimize (DISCO) approach which can construct an extremely efficient classifier for low dimensional problems by combining feature selection, efficient discretization, novel bin placement, and lookup. As feature selection and discretization parameters are crucial, appropriate combinatorial optimization is an important aspect of the approach. A performance evaluation is performed for a YouTube classification task using a cellular traffic data set. The initial evaluation results show that the DISCO approach can move the Pareto boundary in the classification performance versus runtime trade-off by up to an order of magnitude compared to runtime optimized random forest and decision tree classifiers.

Automatic and fast detection of overlaid URLs and text in images and video with adult content can be an important aid for children protection in the internet. In this paper, we propose a method and associated algorithms for URL and text extraction from images and video frames based on combined processing in the compressed and spatial domains. It includes an improved approach for text detection and extraction in the discrete cosine transform (DCT) domain. It also includes an adaptive local binarisation with background compensation in the spatial domain to create an image appropriate for optical character recognition (OCR). As a final step, text recognition and verification is performed by matching to an URL dictionary. The proposed algorithms are quantitatively and qualitatively tested against other contemporary techniques, and encouraging results are obtained. Experimental tests over 3,000 images have been conducted highlighting the efficiency of the proposed approach.