The THINGS images database is a freely available database of tens of thousands labelled images sorted into 1854 "concepts" (http://doi.org/10.17605/osf.io/jum2f). Relatively few technical details relevant to studies in cognitive neuroscience have been described. The aim of this project is to provide a summary of variation in luminance and luminance contrast within the THINGS database.
All analysis code are Matlab files, and all images and data generated by the code are also given here.
In particular, the files THINGS_imStats.m and THINGS_imStats.csv include the statistics of all 26107 images, organised in order of concept (alphabetically), and then in the order of the numbered files within each concept.
THINGS_imageStatistics_v02.m analyses statistics
classification_v01.m classifies images in pairs according to leave one out analysis
classification_allCats_v01.m classifies images into one of 1854 concepts, as opposed to pairs
classification_allStats_v01.m classifies images in pairs using spatial frequency and oriented energy in addition to mean luminance and RMS contrast
normColIms.m shows an example of how to normalise the luminance and RMS contrast of coloured images
THINGS_imageStatistics_v01_normedIms_v01.m repeats the pairwise classification, but for normalised images
Data figures in the manuscript were created with DataGraph.