In this paper, we present a study on skeletonization of real-world shape data. The data stem from the cultural heritage domain and represent contact tracings of prehistoric petroglyphs. Automated analysis can support the work of archeologists on the investigation and categorization of petroglyphs. One strategy to describe petroglyph shapes is skeletonbased. The skeletonization of petroglyphs is challenging since their shapes are complex, contain numerous holes and are often incomplete or disconnected. Thus they pose an interesting testbed for skeletonization. We present a large real-world dataset consisting of more than 1100 petroglyph shapes. We investigate their properties and requirements for the purpose of skeletonization, and evaluate the applicability of state-of-the-art skeletonization and skeleton pruning algorithms on this type of data. Experiments show that pre-processing of the shapes is crucial to obtain robust skeletons. We propose an adaptive pre-processing method for petroglyph shapes and improve several state-of-the-art skeletonization algorithms to make them suitable for the complex material. Evaluations on our dataset show that 79.8 % of all shapes can be improved by the proposed pre-processing techniques and are thus better suited for subsequent skeletonization. Furthermore we observe that a thinning of the shapes produces robust skeletons for 83.5 % of our shapes and outperforms more sophisticated skeletonization techniques.
This paper describes the development, implementation and evaluation of a speech/music detector. We aim at audio from different sources with different qualities -i.e. audio from "the wild". We examine existing approaches for audio classification and select a recent feature. We modify the feature and evaluate the classification accuracy on a random test set of more than 60 hours of audio material against a standard speech/music detection approach. With our approach, we reach a classification accuracy of 96,6%. We provide a performant open source implementation of our detector.
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.