Colour - Datasets

Colour science datasets for use with Colour or any Python package manipulating colours. The datasets are hosted in Zenodo under the Colour Science - Datasets community.

It is open source and freely available under the BSD-3-Clause terms.

1   Features

Colour - Datasets was created to overcome issues encountered frequently when trying to access or use colour science datasets:

• No straightforward ingestion path for dataset content.
• No simple loading mechanism for dataset content.
• Unavailability of the dataset, e.g. download url is down, dataset content is passed directly from hand to hand.
• No information regarding the definitive origination of the dataset.

Colour - Datasets offers all the above: it allows users to ingest and load colour science datasets with a single function call. The datasets information is hosted on Zenodo where the record for a dataset typically contain:

• An urls.txt file describing the urls to source the dataset files from.
• A copy of those files in the eventuality where the source files are not available or the content has changed without notice.
• Information about the authors, content and licensing.

When no explicit licensing information is available, the dataset adopts the Other (Not Open) licensing scheme, implying that assessing usage conditions is at the sole discretion of the users.

1.1   Examples

Colour - Datasets can also be used online with Google Colab.

Most of the objects are available from the colour_datasets namespace:

import colour_datasets

The available datasets are listed with the colour_datasets.datasets() definition:

print(colour_datasets.datasets())

colour-science-datasets
=======================

Datasets : 23
Synced   : 1
URL      : https://zenodo.org/communities/colour-science-datasets/

Datasets
--------

[ ] 3269926 : Agfa IT8.7/2 Set - Marszalec (n.d.)
[ ] 8314702 : Camera Dataset - Solomatov and Akkaynak (2023)
[ ] 3245883 : Camera Spectral Sensitivity Database - Jiang et al. (2013)
[ ] 3367463 : Constant Hue Loci Data - Hung and Berns (1995)
[ ] 3362536 : Constant Perceived-Hue Data - Ebner and Fairchild (1998)
[ ] 3270903 : Corresponding-Colour Datasets - Luo and Rhodes (1999)
[ ] 3269920 : Forest Colors - Jaaskelainen et al. (1994)
[ ] 4394536 : LUTCHI Colour Appearance Data - Luo and Rhodes (1997)
[x] 3245875 : Labsphere SRS-99-020 - Labsphere (2019)
[ ] 3269924 : Lumber Spectra - Hiltunen (n.d.)
[ ] 4051012 : Measured Commercial LED Spectra - Brendel (2020)
[ ] 3269918 : Munsell Colors Glossy (All) (Spectrofotometer Measured) - Orava (n.d.)
[ ] 3269916 : Munsell Colors Glossy (Spectrofotometer Measured) - Haanpalo (n.d.)
[ ] 3269914 : Munsell Colors Matt (AOTF Measured) - Hauta-Kasari (n.d.)
[ ] 3269912 : Munsell Colors Matt (Spectrofotometer Measured) - Hauta-Kasari (n.d.)
[ ] 3245895 : New Color Specifications for ColorChecker SG and Classic Charts - X-Rite (2016)
[ ] 3252742 : Observer Function Database - Asano (2015)
[ ] 3269922 : Paper Spectra - Haanpalo (n.d.)
[ ] 6590768 : Physlight - Camera Spectral Sensitivity Curves - Winquist et al. (2022)
[ ] 3372171 : RAW to ACES Utility Data - Dyer et al. (2017)
[ ] 4642271 : Spectral Database of Commonly Used Cine Lighting - Karge et al. (2015)
[ ] 4297288 : Spectral Sensitivity Database - Zhao et al. (2009)
[ ] 4050598 : Spectral Upsampling Coefficient Tables - Jakob and Hanika. (2019)

A ticked checkbox means that the particular dataset has been synced locally. A dataset is loaded by using its unique number: 3245895:

print(colour_datasets.load("3245895").keys())

Pulling "New Color Specifications for ColorChecker SG and Classic Charts" record content...
Downloading "urls.txt" file: 8.19kB [00:01, 5.05kB/s]
Downloading "ColorChecker24_After_Nov2014.zip" file: 8.19kB [00:01, 6.52kB/s]
Downloading "ColorChecker24_Before_Nov2014.zip" file: 8.19kB [00:01, 7.66kB/s]
Downloading "ColorCheckerSG_After_Nov2014.zip" file: 8.19kB [00:01, 7.62kB/s]
Downloading "ColorCheckerSG_Before_Nov2014.zip" file: 8.19kB [00:00, 9.39kB/s]
Unpacking "/Users/kelsolaar/.colour-science/colour-datasets/3245895/dataset/ColorCheckerSG_Before_Nov2014.zip" archive...
Unpacking "/Users/kelsolaar/.colour-science/colour-datasets/3245895/dataset/ColorCheckerSG_After_Nov2014.zip" archive...
Unpacking "/Users/kelsolaar/.colour-science/colour-datasets/3245895/dataset/ColorChecker24_After_Nov2014.zip" archive...
Unpacking "/Users/kelsolaar/.colour-science/colour-datasets/3245895/dataset/ColorChecker24_Before_Nov2014.zip" archive...
odict_keys(['ColorChecker24 - After November 2014', 'ColorChecker24 - Before November 2014', 'ColorCheckerSG - After November 2014', 'ColorCheckerSG - Before November 2014'])

Alternatively, a dataset can be loaded by using its full title: New Color Specifications for ColorChecker SG and Classic Chart - X-Rite (2016)s

print(colour_datasets.load("3245895").keys())

odict_keys(['ColorChecker24 - After November 2014', 'ColorChecker24 - Before November 2014', 'ColorCheckerSG - After November 2014', 'ColorCheckerSG - Before November 2014'])

2   User Guide

2.1   Installation

2.1.1   Primary Dependencies

Colour - Datasets requires various dependencies in order to run:

• python >= 3.9, < 4
• cachetools
• colour-science >= 4.3
• imageio >= 2, < 3
• numpy >= 1.22, < 2
• scipy >= 1.8, < 2
• tqdm
• xlrd

2.1.2   Pypi

Once the dependencies are satisfied, Colour - Datasets can be installed from the Python Package Index by issuing this command in a shell:

pip install --user colour-datasets

The overall development dependencies are installed as follows:

pip install --user 'colour-datasets[development]'

2.2   Contributing

If you would like to contribute to Colour - Datasets, please refer to the following Contributing guide for Colour.

2.3   Bibliography

The bibliography is available in the repository in BibTeX format.

3   API Reference

The main technical reference for Colour - Datasets is the API Reference.

4   Code of Conduct

The Code of Conduct, adapted from the Contributor Covenant 1.4, is available on the Code of Conduct page.

5   Contact & Social

The Colour Developers can be reached via different means:

• Email
• Facebook
• Github Discussions
• Gitter
• Twitter

6   About