|
Convolutional Neural Networks (CNNs) have delivered impressive state-of-the-art performances for many vision tasks, while the computation costs of these networks during test-time are notorious. Empirical results have discovered that CNNs have learned the redundant representations both within and across different layers. When CNNs are applied for binary classification, this article investigate a method to exploit this redundancy across layers, and construct a cascade of classifiers which explicitly balances classification accuracy and hierarchical feature extraction costs.%Rather than reducing redundancy within convolution filters,This method cost-sensitively selects feature points across several layers from trained networks and embeds non-expensive yet discriminative features into a cascade. Experiments on binary classification demonstrate that our framework leads to drastic test-time improvements, e.g., possible $47.2 imes$ speedup for TRECVID upper body detection, $2.82 imes$ speedup for Pascal VOC2007 People detection, $3.72 imes$ for INRIA Person detection % and average $2.5 imes$ speedup for CIFAR-10with less than 0.5% drop in accuracies of the original networks. |
|
Keywords:Convolutional Neural Network, Pattern Recognition, Binary Classification, Cost-Sensitive Classification, Feature Selection, Cascaded Classifier, Image Classification |
|