THINGS: A database of 1,854 object concepts and more than 26,000 naturalistic object images

Abstract

In recent years, the use of a large number of object concepts and naturalistic object images has been growing enormously in cognitive neuroscience research. Classical databases of object concepts are based mostly on a manually-curated set of concepts. Further, databases of naturalistic object images typically consist of single images of objects cropped from their background, or a large number of uncontrolled naturalistic images of varying quality, requiring elaborate manual image curation. Here we provide a set of 1,854 diverse object concepts sampled systematically from concrete picturable and nameable nouns in the American English language. Using these object concepts, we conducted a large-scale web image search to compile a database of 26,107 high-quality naturalistic images of those objects, with 12 or more object images per concept and all images cropped to square size. Using crowdsourcing, we provide higher-level category membership for the 27 most common categories and validate them by relating them to representations in a semantic embedding derived from large text corpora. Finally, by feeding images through a deep convolutional neural network, we demonstrate that they exhibit high selectivity for different object concepts, while at the same time preserving variability of different object images within each concept. Together, the THINGS database provides a rich resource of object concepts and object images and offers a tool for both systematic and large-scale naturalistic research in the fields of psychology, neuroscience, and computer science.

Revealing the multidimensional mental representations of natural objects underlying human similarity judgments

“…First, we needed to identify a set of objects that is representative of the objects encountered in the real world. For that purpose, we chose the 1,854 objects in the THINGS database 17 , which we developed to provide a comprehensive list of living and non-living things according to their everyday use in the American English language. For each object, we chose a representative image that had been shown to be named consistently during the creation of this database.…”

“…Task and modeling procedure for large-scale identification of mental object representations. a We applied a triplet odd-one-out similarity task to images of the 1,854 objects in the THINGS database 17 and collected a large number of ratings (1.46 million) using online crowdsourcing. The triplet odd-one-out task measures object similarity as the probability of choosing two objects together.…”

“…This indicates that categorization behavior for many categories may be accounted for by the similarity of objects, a property which has been discussed previously 28,29 but which had not been tested on a large set of natural objects. To test how well natural categories could be predicted by the similarity embedding, we used 18 unique high-level categories identified in the THINGS database 17 and used a cross-validated nearest centroid classifier to predict category membership for each of the 1,112 objects of those categories. The classifier performed at 86.42% (chance performance: 5.56%), on par with a recent semantic embedding of object meaning 30 (85.97%) that had been trained on billions of words, demonstrating that the dimensions we identified allow the prediction of categorization behavior for a large number of natural categories.…”

A Study of Representational Similarity: The Emergence of Object Concepts in Rapid Serial Visual Presentation Streams

“…The present study used the THINGS (Hebart et al, 2019) database. The database is comprised of 1,854 of common concrete object concepts that were systematically sampled from the American English language.…”

“…The selection process for the candidate concepts and images was rigorous. Concepts and their categories were validated by the workers from Amazon Mechanical Turk and the WordNet word-sense disambiguation (Hebart et al, 2019). These images were also fed through the layers in deep convolutional neural network CorNet-S (Kubilius et al, 2018) to ensure that they retain a sufficient degree of visual variability.…”