The Laval Indoor HDR Dataset
Full-HDR high resolution indoor panoramas
This dataset contains 2100+ high resolution indoor panoramas, captured using a Canon 5D Mark III and a robotic panoramic tripod head. Each capture was multi-exposed (22 f-stops) and is fully HDR, without any saturation. Panoramas were stitched from 6 captures (60 degrees azimuth increment) and were captured in a wide variety of indoor environments.
To obtain the dataset, you can:
- Request full access to the dataset if you are an academic researcher. First fill the End User Licence Agreement and return it to Jean-Francois Lalonde by email at jflalonde at gel dot ulaval dot ca. For other uses, see Usage below.
- Download a ZIP archive containing a preview dataset (100 samples, 2048x1024, full-HDR). Note that the usage of this preview dataset is restricted to academic or non-profit researchers (see Usage).
- Consult this preview where all the dataset is exposed in a single page. The data is however tonemapped (LDR), of limited resolution (512x256), and JPEG compressed, so it should only be used as an overview.
This data is available for general use by:
- Non-profit organisation (for academic or government-sponsored researchers): this license does not grant the rights to use the data set or any derivation of it for commercial activities neither for redistribution. This license is free.
- For-profit organisation: this license allows the access to the data, the use of the data to create or improve models and resulting output, with the right to make the output available to third parties or to use if for your benefit. The dataset itself cannot be made available to third party. For any commercial use please contact Jean-Francois Lalonde at jflalonde at gel dot ulaval dot ca.
If you use this data in a publication, we ask you to cite this paper:
The authors gratefully acknowledge Adobe for funding this project. We also thank Jean-Michel Fortin, Samuel Delisle, Guillaume Doyon, and Gabriel Lavin-Muller for their help with data capture and processing. Thanks also to Dominic Bilodeau, Louis-Emile Robitaille and Frédéric St-Pierre who participated in earlier iterations of the data capture process.