This paper proposes a robust classification framework, which brings together effective normalization measures, visible features, and image attributes to assemble a helpful system. The overview of the proposed framework is proven in Figure 1. There are three primary approaches with emphasis on reducing coaching samples and bettering the effectivity of learning machines. First, PSM is used to extend the training knowledge area, which allows us to generate ideal robust classifiers with out having to collect a large quantity of training samples.
