Adapting RANSAC SVM to Detect Outliers for Robust Classification

Most visual classification tasks assume the authenticity of the label information. However, due to several reasons such as difficulty of annotation or inadvertently due to human error, the annotation can often be noisy. This results in wrongly annotated examples. In this paper, we consider the examples that are wrongly annotated to be outliers. The task of learning a robust inlier model in the presence of outliers is typically done through the RANSAC algorithm[6]. We show that instead of adopting RANSAC to obtain the ‘right’ model, we could use many instances of randomly sampled subsets to build a lot of models. The collective decision of all these models can be used to identify examples that are likely to be outliers. This results in a modification to RANSAC SVM[11] to explicitly obtain probable outliers from the set of given examples. Once the outliers are detected, these examples are excluded from the training set. We also show that the method can be used to identify very hard examples present in the training data. In this case, where we believe that the examples are correctly annotated, we can achieve good generalization when such examples are excluded from the training set. The method is evaluated using the standard PASCAL VOC 2007 dataset[4]. We show that the method is particularly suited for identifying wrongly annotated examples resulting in improvement of more than 12% over the RANSAC-SVM approach. Hard examples in PASCAL VOC dataset are also identified by this method and this even results in a marginal improvement of the mean average precision over the base classifier provided with all clean examples.