Context A labeled point-cloud dataset taken from the Semantic3D project (http://semantic3d.net/view_dbase.php?chl=1). The dataset has billions of XYZ-RGB points and labels them into 7 classes. Content The data are raw ASCII files containing 7 columns (X, Y, Z, Intensity, R, G, B) and the labels are `{1: man-made terrain, 2: natural terrain, 3: high vegetation, 4: low vegetation, 5: buildings, 6: hard scape, 7: scanning artefacts, 8: cars}` including an 8th class of unlabeled. Acknowledgements The data are taken directly from the Semantic3D competition and users must check and cite the rules and regulations posted on the original site: http://semantic3d.net/view_dbase.php?chl=1 Inspiration 1. What sort of models can classify point clouds well? 2. What transformations make classification easier? 3. Are there certain classes which require more data in order to classify well?