nielsr's picture
nielsr HF Staff
Add dataset card and metadata
e7ebf6e verified
|
raw
history blame
2.07 kB
metadata
task_categories:
  - image-to-image

MobileSpectralCCDataset

This dataset accompanies the paper Leveraging Multispectral Sensors for Color Correction in Mobile Cameras, presented at CVPR 2026.

Project Page | GitHub Repository

Overview

The MobileSpectralCCDataset is a physically grounded synthetic dataset designed to support end-to-end color correction research using auxiliary multispectral (MS) sensors. It was constructed by aggregating and repurposing hyperspectral reflectance data from two publicly available datasets: KAUST and BJTU-UVA.

Key Features:

  • Sensor Simulation: Includes simulated high-resolution RGB and low-resolution multispectral measurements across a wide range of illuminants and camera spectral sensitivities.
  • Ground Truth: Color references are rendered under the standard D65 illuminant.
  • Geometric Inconsistency: Includes a misaligned version of the data featuring spatial offsets and realistic warping transformations (derived from the Zurich dataset) to mimic real-world dual-sensor system alignments.

Dataset Structure

The dataset consists of pairs of high-resolution RGB images and auxiliary low-resolution MS images, along with their corresponding ground-truth color-corrected versions. For more details on the generation process, please refer to the official GitHub repository.

Citation

If you use this dataset in your research, please cite:

@inproceedings{leveraging2026cogo,
  title={Leveraging Multispectral Sensors for Color Correction in Mobile Cameras}, 
  author={Luca Cogo, Marco Buzzelli, Simone Bianco, Javier Vazquez-Corral, Raimondo Schettini},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2026}
}