3DUNDERWORLD-SLS
github.com
- Free plan
- Linux
What 3DUNDERWORLD-SLS is
3DUNDERWORLD-SLS is a free Linux project for reconstructing point clouds from image sequences illuminated by patterned light. It is described as an open-source structured-light scanner for rapid geometry acquisition and provides CPU and GPU implementations. Version 4.x expects two or more cameras and offers a generic camera interface that programmers can extend. Camera acquisition is not included, so users must implement a camera class through the supplied interfaces. The command-line program accepts camera image folders and calibration files and exports reconstructed models as PLY or OBJ. Libraries and CPU/GPU examples support integration into other projects, and an API was included in OpenCV 3.1 for a stereo, time-multiplexed Gray-encoding approach. The repository contains demo binaries, example data and CMake build instructions. Dependencies listed are OpenCV 2, CUDA 6.0 or later and glm. CUDA detection determines whether CPU and GPU constructors are built.
Who it suits
It suits Linux developers and researchers building structured-light scanning workflows or integrating reconstruction libraries. Users need multiple cameras and must implement camera acquisition.
Good to know, and what to watch
- Free Linux software
- CPU and GPU implementations
- Exports PLY and OBJ
- Libraries and API for integration
- CMake build script included
- Camera acquisition is not implemented
- Requires two or more cameras
- No real-time capture
- No mesh reconstruction
The verdict
3DUNDERWORLD-SLS provides a programmable point-cloud reconstruction pipeline with CPU and optional CUDA GPU paths. It is most appropriate when developers can supply camera acquisition and work within its older listed dependencies.
What the maker publishes
- Build
- The repository includes demo binaries and data, and a CMake script to compile the binaries and libraries. source
- Build requirements
- The listed dependencies are OpenCV 2, CUDA 6.0 or later, and glm. source
- Camera acquisition
- Camera acquisition is not implemented, though the project provides interfaces for users to implement a camera class. source
- Camera limitation
- Camera acquisition is not implemented; the project provides interfaces for users to implement a camera class. source
- CPU and GPU
- The reconstruction process is implemented for both CPU and GPU, with CUDA-enabled builds generating both CPU and GPU executables when CUDA is detected. source
- Data size
- The repository notes that its included example data makes the repository about 300 MB. source
- Dependencies
- The listed dependencies are OpenCV 2, CUDA 6.0 or later, and glm. source
- Download and build
- The repository includes demo binaries and data and provides CMake instructions to build the binaries and libraries. source
- Founded
- 2011 source
- GPU support
- If CUDA is detected, the default build creates CPU and GPU constructors; otherwise, it generates only the CPU constructor. source
- Headquarters
- Montreal, Quebec source
- Integration
- An API for 3DUNDERWORLD-SLS was included in OpenCV 3.1 and implements time-multiplexing Gray encoding for a stereo approach. source
- Integration
- The project provides libraries intended for integration into other projects, with CPU and GPU application examples in the repository. source
- Intended use
- The project's technical report describes the system as an open-source structured-light scanner for rapid geometry acquisition. source
- Intended users
- The lab describes its work as research and notes collaborations with industry and other external partners. source
- Library
- The project includes libraries intended for integration into other projects, with CPU and GPU application examples. source
- License
- The repository includes the GNU Lesser General Public License version 3 with additional permissions. source
- License
- The GitHub repository identifies its license as LGPL-3.0. source
- Output
- The command-line program can export reconstructed models in PLY or OBJ format. source
- Outputs
- The demo accepts camera image folders and calibration files and can save output as PLY or OBJ. source
- Processing
- The project provides CPU and GPU implementations of the reconstruction process. source
- Processing
- Image processors use images, projector parameters, and camera calibration parameters to produce data that the reconstructor combines into a point cloud. source
- Project status
- The lab's software page lists GPU-based structured-light scanning software V4.0 under “Under Active Development.” source
- Publication requirement
- The repository says users must cite specified publications in any resulting publication. source
Checked against the maker's pages on 8 Oct 2026.