Visual Genome Image Semantic Segmentation Dataset

The Visual Genome dataset is a machine vision dataset that contains 108,077 images, 5.4 million region descriptions, 1.7 million visual question-and-answer data, 3.8 million target instances, 2.8 million attributes, and 2.3 million relationship information. The dataset was jointly completed by Stanford University, Yahoo, Toyota, etc., and is committed to closely connecting image information with language.