The ICL-NUIM dataset aims at benchmarking RGB-D, Visual Odometry and SLAM algorithms. Two different scenes (the living room and
the office room scene) are provided with ground truth. Living room has 3D surface ground truth together with the depth-maps as
well as camera poses and as a result pefectly suits not just for bechmarking camera trajectory but also reconstruction. Office
room scene comes with only trajectory data and does not have any explicit 3D model with it. Below we provide data for four
different handheld trajectories that are obtained by running Kintinuous on real image data, and finally used in synthetic framework
for obtaining ground truth.
All data is compatible with the evaluation tools available for the TUM RGB-D dataset, and if your system can take TUM RGB-D format PNGs as input, our TUM RGB-D Compatible data will also work (given the correct camera parameters, provided here, while the TUM RGB-D PNG format is described here). We also provide data with synthetic sensor noise, as described in our paper.
Also, please visit SceneNet for more 3D models (but without any colour yet).
This work is in collaboration with Thomas Whelan and John McDonald from National University of Ireland, Maynooth. The ICRA paper can be download from this link. If you find our work useful, please use the following bibtex entry to cite the paper.
The data is released under Creative Commons 3.0 (CC BY 3.0), see http://creativecommons.org/licenses/by/3.0/.
All data is compatible with the evaluation tools available for the TUM RGB-D dataset, and if your system can take TUM RGB-D format PNGs as input, our TUM RGB-D Compatible data will also work (given the correct camera parameters, provided here, while the TUM RGB-D PNG format is described here). We also provide data with synthetic sensor noise, as described in our paper.
Also, please visit SceneNet for more 3D models (but without any colour yet).
| Trajectories for 3D reconstruction: Living Room Dataset (click on the link for more) |
|
Living Room 'lr kt0' Number of Images: 1510. Size: 3.3G Frame Rate: 30Hz Total Time: 51 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT Global Poses [R | t]: Global_RT_Trajectory_GT |
Living Room 'lr kt1' Number of Images: 967. Size: 2.1G Frame Rate: 30Hz Total Time: 33 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT Global Poses [R | t]: Global_RT_Trajectory_GT |
|---|---|
|
Living Room 'lr kt2' Number of Images: 882. Size: 1.9G Frame Rate: 30Hz Total Time: 30 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT Global Poses [R | t]: Global_RT_Trajectory_GT |
Living Room 'lr kt3' Number of Images: 1242. Size: 2.6G Frame Rate: 30Hz Total Time: 42 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT Global Poses [R | t]: Global_RT_Trajectory_GT |
| Surface Reconstruction Quality Evaluation |
 |
Reconstructions produced by multi-view stereo fusion approaches or RGB-D dense fusion methods can be compared against this ground truth 3D surface model using our SurfReg tool, available here. |
| Trajectories for Visual Odometry: Office Room Dataset (click on the link for more) |
|
Office Room 'of kt0' Number of Images: 1510. Size: 2.9G Frame Rate: 30Hz Total Time: 51 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT |
Office Room 'of kt1' Number of Images: 967. Size: 1.7G Frame Rate: 30Hz Total Time: 33 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT |
|---|---|
|
Office Room 'of kt2' Number of Images: 882. Size: 1.7G Frame Rate: 30Hz Total Time: 30 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT |
Office Room 'of kt3' Number of Images: 1242. Size: 2.3G Frame Rate: 30Hz Total Time: 42 secs ICL-NUIM PNGs TUM RGB-D Compatible PNGs TUM RGB-D Compatible PNGs with noise Poses TUM RGB-D format: TrajectoryGT |
| Authors |
This work is in collaboration with Thomas Whelan and John McDonald from National University of Ireland, Maynooth. The ICRA paper can be download from this link. If you find our work useful, please use the following bibtex entry to cite the paper.
@InProceedings{handa:etal:ICRA2014,
author = {A. Handa and T. Whelan and J.B. McDonald and A.J. Davison},
title = {A Benchmark for {RGB-D} Visual Odometry, {3D} Reconstruction and {SLAM}},
booktitle = {IEEE Intl. Conf. on Robotics and Automation, ICRA},
address = {Hong Kong, China},
month = {May},
year = {2014}
}
|
| License |
The data is released under Creative Commons 3.0 (CC BY 3.0), see http://creativecommons.org/licenses/by/3.0/.