2 datasets found

The dataset used in the paper is a set of images, and the authors used it to train and test their ladder network model.

The Visual Genome dataset is a large-scale visual question answering dataset, containing 1.5 million images, each with 15-30 annotated entities, attributes, and relationships.