LidarScout: Direct Out-of-Core Rendering of Massive Point Clouds

LidarScout: Direct Out-of-Core Rendering of Massive Point Clouds

This dataset contains data related to our LidarScout paper at High-Performance Graphics 2025. This paper proposes a method for instantly exploring huge, compressed point clouds (LAZ) without any pre-processing and no extra storage. Contained data is:

• LidarScout executable for Windows (requires CUDA)
• A screen-capture video as MP4
• Source code for the viewer (Github Snapshot from 2025-09-25, https://github.com/cg-tuwien/lidarscout, mostly C++ and CUDA)
• Source code for the training (Github Snapshot from 2025-09-25, https://github.com/cg-tuwien/lidarscout_training, mostly Python)
• Trained models for all variants as Pytorch checkpoints (.ckpt) and Torchscript (.pt)
• Training and testing datasets based on AHN5, Bund_BoraPk, CA13, ID15_Bunds, NZ23_Gisborne, BR17_SaoPaulo, and SwissSurface3D as CSV, PLY, and BIN files.
• Testing results for all model variants and all test datasets as PLY and NPY files

More detailed information on the folder and file structure can be found in the README of the training repository.

More general information: https://www.cg.tuwien.ac.at/research/publications/2025/erler-2025-lidarscout/

There are some scripts provided in the code archive for very interested users; they may require a little bit of tinkering to get them to run.

Licenses

The data is licensed under CC-BY 4.0, the code is licensed under MIT.

Acknowledgements

We thank the following data set providers: Bunds at el. and Open Topography for the Bund_Bora [ BDG+19] and ID15_Bunds [BDG+20] data sets; PG&E and Open Topography for CA13 [ Pac13 ]; The Ministry of Business, Innovation and Employment and Toit¯u Te Whenua Land Information New Zealand and Open Topography for Gisborne [ MoBE24 ]; The São Paulo City Hall (PMSP) and Open Topography for São Paulo [ (PM17 ]; The Bundesamt für Landestopografie swisstopo for swissSURFACE3D [Swi20].

We thank Paul Guerrero, Pedro Hermosilla, and Adam Celarek for their valuable inputs. Further, we thank Stefan Ohrhallinger for running reconstructions with BallMerge [POEM24].
This research has been funded by WWTF project ICT22-055 - Instant Visualization and Interaction for Large Point Clouds.