Attribute classification for transformer substation based on deep convolutional network

Abstract

Search engines of transformer substation receive a large number of category-related and content-based queries. These queries are best answered by the attribute listings, which contain the semantic object information. While creating such listing needs the requirement analysis to figure out the most concerned attributes. In this paper, we build a new dataset of transformer substation for attribute classification. To efficiently annotate the image set, we employ the deep convolutional network, which outputs the posterior probability of each attribute. The extensive experiments show that our work reaches 94.67% for attribute classification, which demonstrate the efficiency of proposed work.