Sensitivity Analysis of Neural Network Parameters for Advertising Images Detection
Changsheng Gong, Fuxi Zhu, Haitao Yao · 2010
Web advertising has become a major industry and many advertisements appear in the form of images. Although it makes considerable profit, these advertisements tend to disturb the internet surfing of normal users. Moreover, they always bring extra burden in indexing to commercial image search engines. In this paper we present the development and performance of a Neural Network (NN) for advertising images detection as well as the results of applying sensitivity analysis on the NN parameters in order to identify the factors that results a web image into an advertising one. A sample of 1262 web images was used to train and test our NN. Ad word count in image URL and Does image's URL point to outside site were identified as the two factors most contributory to advertising images detection.