Building effective SVM concept detectors from clickthrough data for large-scale image retrieval

Abstract

Clickthrough data is a source of information that can be used for automatically building concept detectors for image retrieval. Previous studies, however, have shown that in many cases the resulting training sets suffer from severe label noise that has a significant impact in the SVM concept detector performance. This paper evaluates and proposes a set of strategies for automatically building effective concept detectors from clickthrough data. These strategies focus on: (1) automatic training set generation; (2) assignment of label confidence weights to the training samples and (3) using these weights at the classifier level to improve concept detector effectiveness. For training set selection and in order to assign weights to individual training samples three Information Retrieval (IR) models are examined: vector space models, BM25 and language models. Three SVM variants that take into account importance at the classifier level are evaluated and compared to the standard SVM: the Fuzzy SVM, the Power SVM, and the Bilateral-weighted Fuzzy SVM. Experiments conducted on the MM Grand Challenge dataset (consisting of 1M images and 82.3M unique clicks) for 40 concepts demonstrate that (1) on average, all weighted SVM variants are more effective than the standard SVM; (2) the vector space model produces the best training sets and best weights; (3) the Bilateral-weighted Fuzzy SVM produces the best results but is very sensitive to weight assignment and (4) the Fuzzy SVM is the most robust training approach for varying levels of label noise.