Cross-domain few-shot classification task (CD-FSC) combines few-shot classification with the requirement to generalize across domains represented by datasets. This setup faces challenges originating from the limited labeled data in each class and, additionally, from the domain shift between training and test sets. In this paper, we introduce a novel training approach for existing FSC models. It leverages on the explanation scores, obtained from existing explanation methods when applied to the predictions of FSC models, computed for intermediate feature maps of the models. Firstly, we tailor the layer-wise relevance propagation (LRP) method to explain the prediction outcomes of FSC models. Secondly, we develop a model-agnostic explanation-guided training strategy that dynamically finds and emphasizes the features which are important for the predictions. Our contribution does not target a novel explanation method but lies in a novel application of explanations for the training phase. We show that explanation-guided training effectively improves the model generalization. We observe improved accuracy for three different FSC models: RelationNet, cross attention network, and a graph neural network-based formulation, on five few-shot learning datasets: miniImagenet, CUB, Cars, Places, and Plantae.
High-resolution PolSAR images are wildly used in land cover mapping. However, there are two problems when working with large areas, i.e., long processing time and large computer memory. Considering that properties of homogeneous area, such as forest, farm land, bare land, etc., are almost the same under different resolutions, which means that these areas can be processed in relatively low-resolution without decreasing mapping accuracy, this paper proposes a new land cover mapping method, to shorten processing time and reduce the demand of memory when high resolution images are used. In this method, relatively low-resolution images, used for homogeneous areas mapping, are obtained via pyramid transformation from the original image, due to the fact that pyramid transformation has good performance in retaining main information in the transformed images. The traditional pyramid transformation is adapted to PolSAR images. To verify the proposed method, iterated Freeman-Wishart classification is used based on pyramid transformation in this paper. Mapping accuracy and processing time of classification based on pyramid transformation and without pyramid transformation are compared. Experimental results show that processing time of classification based on pyramid transformation is reduced dramatically and the classification accuracy is improved.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.