Broden

• Broadly and Densely Labeled Dataset
• unifies several densely labeled image data sets: ADE [43], OpenSurfaces [4], Pascal-Context [19], Pascal-Part [6], and the Describable Textures Dataset [7]
• These data sets contain examples of a broad range of objects, scenes, object parts, textures, and materials in a variety of contexts
• segmented down to the pixel level except textures and scenes which are given for full-images
• every image pixel in the data set is annotated with one of the eleven common color names according to the human perceptions classified by van de Weijer
• The concept labels in Broden are normalized and merged from their original data sets so that every class corresponds to an English word
• Labels are merged based on shared synonyms, disregarding positional distinctions such as ‘left’ and ‘top’