SlideImages: A Dataset for Educational Image Classification

摘要: In the past few years, convolutional neural networks (CNNs) have achieved impressive results in computer vision tasks, which however mainly focus on photos with natural scene content. Besides, non-sensor derived images such as illustrations, data visualizations, figures, etc. are typically used to convey complex information or explore large datasets. However, this kind of has received little attention vision. CNNs and similar techniques use volumes training data. Currently, many document analysis systems trained part due lack datasets educational image paper, we address issue present SlideImages, a dataset for task classifying illustrations. SlideImages contains collected from various sources, e.g., Wikimedia Commons AI2D dataset, test slides. We reserved all actual order ensure that approaches using generalize well new images, potentially other domains. Furthermore, baseline system standard deep architecture discuss dealing challenge limited