The ISPRS scientific initiative Hessigheim 3D (H3D) on semantic segmentation provides benchmark data in terms of labeled laser scanning point clouds and 3D textured meshes. Our first release earlier this year already provided a UAV data set captured at the village of Hessigheim (Germany). Now additional epochs of that configuration are available. Thus, multi-temporal analyses are feasible.
The point cloud features a mean point density of about 800 pts/m² and the oblique imagery used for colorization and 3D mesh texturing realizes a ground sampling distance of about 2-3 cm. This enables the identification of fine-grained structures and represents the state of the art in UAV-based mapping. The respective point clouds are manually labeled into 11 classes and are additionally used to derive labeled textured 3D meshes as an alternative representation.
While our first release provided a data set captured in March 2018, now additional epochs captured with the same high-resolution sensor configuration in November 2018 and March 2019 are available at our benchmark homepage https://ifpwww.ifp.uni-stuttgart.de/benchmark/hessigheim/default.aspx . As fourth epoch, a LiDAR data set captured from a manned aircraft at lower resolution in March 2016 is added, which incorporates typical characteristics of NMA data sets.
Interested researchers can use these state-of-the-art data sets to test their methods and algorithms on semantic segmentation of 3D point clouds and 3D meshes for geospatial applications. Detailed information on the data, the participation and the submission of results for the H3D benchmark is also available in the accompanying paper https://www.sciencedirect.com/science/article/pii/S2667393221000016?via%3Dihub .
|Contact||Prof. Norbert Haala|