ImageBeagle is a novel tool that efficiently searches through 30M images to find images that maximize a given objective function.

ImageBeagle Dataset: We scraped the internet and publicly available datasets to curate our 30M natural image dataset. In addition to this, we created artificial stimuli that are often used in neuroscience experiments such as gratings, colorful dots, bars...

ImageBeagle Search:  We developed ImageBeagle as a simple search method that alternates between nearest neighbors and global search over coresets. Through this, ImageBeagle can efficiently search through the large image manifold to satisfy a given objective function.