Malcolm Mielle

Recent Publications

See 10 more →

Datasets

3 min read

Datasets are one of the most valuable research outputs — they outlive papers and enable others to build on your work. If you use one of mine, please cite the DOI for the version you used.

SEAR Dataset

Vsevolod Skorokhodov, Chenghao Xu, Shuo Sun, Olga Fink, Malcolm Mielle (2026)

9 scenes (~2,000 images) of paired RGB/thermal trajectories captured under varying illumination and viewpoints, with distinct capture conditions per modality — a benchmark for multimodal 3D scene reconstruction. Introduced in SEAR.

DOI | Code

@dataset{skorokhodov_2026_19057885,
  author = {Skorokhodov, Vsevolod and Xu, Chenghao and Sun, Shuo and Fink, Olga and Mielle, Malcolm},
  title = {{SEAR: RGB+Thermal 3D Reconstruction Dataset}},
  year = 2026,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.19057885},
  url = {https://doi.org/10.5281/zenodo.19057885}
}

ThermoScenes (ThermoNeRF)

Mariam Hassan, Florent Forest, Olga Fink, Malcolm Mielle (2024)

A dataset of paired RGB+thermal images: 8 scenes of building facades and 8 scenes of everyday objects, enabling evaluation of multimodal novel view synthesis in diverse scenarios. Introduced in thermonerf.

DOI | URL

@dataset{hassan_2024_10835108,
  author = {Hassan, Mariam and Forest, Florent and Fink, Olga and Mielle, Malcolm},
  title = {{ThermoNeRF: Multimodal Neural Radiance Fields for Thermal Novel View Synthesis}},
  year = 2024,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.10835108},
  url = {https://zenodo.org/record/10835108}
}

BuildNet3D Dataset

Chenghao Xu, Malcolm Mielle, Antoine Laborde, Ali Waseem, Florent Forest, Olga Fink (2025)

Synthetic and real-world building data for evaluating 3D reconstruction and geometric characteristic estimation (window-to-wall ratio, footprint area) — the benchmark used in buildnet3d.

URL

Ground truth matches between maps with high disparity

Malcolm Mielle, Martin Magnusson, Achim J. Lilienthal (2019)

Ground-truth correspondence data between maps with high disparity — the evaluation data for sketch-map matching from my PhD.

DOI | URL

@dataset{mielle_2019_2574036,
  author = {Mielle, Malcolm and Magnusson, Martin and Lilienthal, Achim J.},
  title = {{Ground truth matches between maps with high disparity}},
  year = 2019,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.2574036},
  url = {https://zenodo.org/record/2574036}
}

SLAM datasets — Radar, Velodyne (2018)

Three public datasets with synchronized radar and Velodyne lidar scans for localization and mapping, collected during my PhD:

DatasetContentsDOI
Dortmund SLAM datasetRadar + Velodyne, urban environment10.5281/zenodo.1489911
Hannover University SLAM datasetRadar + Velodyne, university campus10.5281/zenodo.1489924
Örebro University basement SLAM datasetRadar + Velodyne, indoor basement10.5281/zenodo.1489896

All released under CC-BY 4.0.

@dataset{mielle_2018_1489911,
  author = {Mielle, Malcolm and Magnusson, Martin and Lilienthal, Achim J.},
  title = {{Dortmund SLAM dataset - Radar, Velodyne}},
  year = 2018,
  publisher = {Zenodo},
  doi = {10.5281/zenodo.1489911},
  url = {https://zenodo.org/record/1489911}
}

For the Hannover University and Örebro basement datasets, replace the title, DOI, and URL above with the values from the corresponding table row.

If you use any of these datasets, please also cite the associated paper — see All publications. For more on why citing datasets and software with DOIs matters (and how link rot breaks science), see my note Cite Software and Datasets in Your Research.

Graph View

Backlinks