The measurement of tree parameters from point cloud data collected with terrestrial (TLS) and personal (PLS) laser scanning devices has increased in importance. When merging the spectral information from a camera with a 3D point cloud, a colorized point cloud can be generated. This information can be used to automatically classify trees with colorful markings (dots, streaks) or to recognize numbers written on the trees. The automatic extraction of marked trees is useful for calculating the parameters of the specific trees meant for harvest, producing models and visualizations of the result of possible stand tending operations or to make finding specific trees in the point cloud of large research plots more practical and less time consuming.

This dataset contains colorized single tree point clouds wtih dots, streaks, and numbers collected with a TLS (RIEGL VZ-400i, Riegl Laser Measurement Systems GmbH, Horn, Austria; with an additional Nikon D850 DSLR camera and attached Nikon AF 14mm f/2.8 D ED fisheye lens) and a PLS system (GeoSLAM ZEB Discovery System combining the GeoSLAM ZEB Horizon with the NCTech iSTAR Pulsar 360-degree camera, GeoSLAM Ltd., Nottingham, UK). The dataset consists of single tree point clouds, stem panorama plots showing the trees surface form 0.5 m up to 3 m height and tables containing the attributes to generate the images in *.csv format. Furthermore, the reference data is provided in a *.xlsx file containing tree id (that can also be found in the names of the tree files), tree species, dbh (diameter at breast height) and whether a tree was marked or not (and if so, specifying the type of marking). Additionally, this dataset contains extracted clusters of the numbers on the trees from the TLS data (in two different point-thickness settings), pre-processed images of stem panorama plots (to show the numbers in higher quality), labelled numbers, numbers from the best training and test split, as well as tables containing the results from our digit recognition with a YOLOv8 model.