Is there a publicly available benchmarking database of classified (labeled) LiDAR data? For example, where each point is classified as ground, vegetation, building, road, water, etc?

The point clouds that I find listed in other questions are not complete classified (or complete validated) point clouds. I have yet to find one where each point contains a label that specifies the type of object from which it originated.

A set of this nature would be very useful as a way to quantitatively measure range finding algorithms. The scholarly articles that I have read typically rely on small private sets or upon qualitative measures.

I would be happy with any source of data, but I prefer airborne LiDAR. Also, I would like a broad range of data including cities with buildings, house, and trees and also pure forests, but again, I would take anything I could get.

In other words, I am looking for a LiDAR point cloud which all points were classified with more than 99.9% accuracy (probably manually classified point clouds?).

  • By 'classified range datasets', I mean data points in 3D space, lidar point clouds, but that have been labeled. That is, each point in the set is classified as ground, building, vegetation, and so on. The datasets in the duplicate question are not labeled. They are just collections of unlabeled points. If I develop an algorithm that, for example, extracts buildings points from a point cloud, I need a labeled point cloud in order to measure its performance.
    – jsp
    Dec 1, 2016 at 14:29
  • I've spent at least a few hours going through the sets. Certainly I could have missed one. But I would expect that few are labeled -- it would be a monumentally tedious task to do by hand. I wanted to make sure I am not missing a known, standard dataset that folks in this area use to test their algorithms -- a ground truth set. For example, the video compression community has a fairly standard set of ground truth video clips that are invaluable for developing and comparing new algorithms.
    – jsp
    Dec 1, 2016 at 14:51
  • for Great Britain the Environment agency has released opendata LiDAR/point clouds environment.data.gov.uk/ds/survey/#/survey?grid=TQ36 now with raw point clouds (there is around 3 Terra-bytes of data available)
    – Mapperz
    Dec 1, 2016 at 15:42
  • @AndreSilva I would be happy with any source, but I would prefer airborne lidar. Also, I would like a broad range of data including cities with buildings, house, and trees and also pure forests, but again, I would take anything I could get.
    – jsp
    Dec 1, 2016 at 15:44
  • @Mapperz Thanks. They don't have classified data, but they do have both DTM's and DSM's. The site doesn't specify how the DTM's are created but they may be useful for identifying ground/vs non-ground points.
    – jsp
    Dec 1, 2016 at 15:54

1 Answer 1


The ISPRS Test Project on Urban Classification, 3D Building Reconstruction and Semantic Labeling, subitem Urban classification and 3D reconstruction have reference data for exactly what you want to do.

Two test sites, each containing several test areas for which reference data are available, are provided for the participants in this project in order to evaluate techniques for the extraction of various urban object classes.

Dataset 1 (Vaihingen/Enz, Germany)

Airborne Laserscanner Data: The test area was covered by altogether 10 strips captured with a Leica ALS50 system. Inside an individual strip the average point density is 4 points/m2.

Dataset 2 (Toronto, Canada)

This data set covers an area of about 1.45 km2 in the central area of the City of Toronto in Canada ... and the Optech’s airborne laserscanner (ALTM-ORION M). ... provided 6 points/m2.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.