neuro datasets

bold5000 - large-scale, slow event-related fMRI dataset at 3T for 4 subjects, each viewing 5,254 images from COCO, ImageNet, and SUN

human connectome project - a map of the structural and functional neural connections within and across individuals

natural scenes dataset - large-scale fMRI dataset at 7T for 8 subjects viewing thousands of color natural scenes from COCO (1000 overlapping with BOLD5000)

ecoset - 1.5m images from 565 basic level categories, chosen to be both (i) frequent in linguistic usage, and (ii) rated by human observers as concrete (e.g. ‘table’ is concrete, ‘romance’ is not).

studyforrest - large-scale fMRI dataset of participants watching the movie Forrest Gump

stimulus datasets

tarrlab datasets - The repository for TarrLab stimulus datasets - Greebles, YUFOs, The Object Databank, Face-Place, etc.

sound datasets

Note: Historically, we maintained pointers to many other image datasets, but with the rise of ImageNet et al., the number of image datasets available online has exploded. Many natural and artificial vision labs maintain and distribute image or sound datasets. Your best bet is to Google for the kinds of images you are looking for.


NeonOpen - open source AI for images to drive engagement


(.zip archive) is a random collection of routines written by me (Michael Tarr) to work with PsychToolBox. The key routine is readdata which makes reading in text files somewhat less painful. Now includes an improved readdata routine (v 1.3). Warning: I have no idea if these work anymore - they are pretty old...