Related Experiment Video
Updated: Jul 24, 2026
14:25
Determining 3D Flow Fields via Multi-camera Light Field Imaging
Published on: March 6, 2013
17.2K
3DLF-Scan dataset: Multi-sensor 3D light-field and structured-light scans of 3D-printed Stanford figures
Bohdan Vodianyk1, Anton Popov2,3, Enrique Nava-Baro4
1Universidad de Málaga, Escuela de Ingenierías Industriales, Dept. Ingeniería Mecánica y Eficiencia Energética, C/ Arquitecto Francisco Penalosa, 6, Málaga, Spain.
Data in Brief
|March 11, 2026
Summary
This study introduces 3DLF-Scan, a novel multi-sensor dataset for 3D reconstruction. It features 3D-printed models captured with light-field cameras and a 3D scanner, ideal for benchmarking computer vision algorithms.
Area of Science:
- Computer Vision
- 3D Reconstruction
- Robotics
Background:
- 3D reconstruction and depth estimation are crucial for various applications.
- Existing datasets often lack multi-sensor integration or controlled capture conditions.
Purpose of the Study:
- To present 3DLF-Scan, a comprehensive multi-sensor dataset for 3D vision research.
- To facilitate the development and benchmarking of novel 3D reconstruction and registration algorithms.
Main Methods:
- Acquisition of 3D-printed Stanford models using light-field cameras and a structured-light 3D scanner.
- Controlled 360° turntable capture with precise pose estimation.
- Inclusion of RGB images, depth maps, masks, point clouds, and calibration files.
Main Results:
- A dataset containing detailed 3D data for canonical models.
- Data organized in standard formats for compatibility with 3D vision toolchains.
- Includes per-view depth completion and metric point clouds.
Conclusions:
- 3DLF-Scan provides a valuable resource for advancing multi-view 3D reconstruction and cross-sensor registration.
- The dataset supports research in depth completion and volumetric fusion techniques.
