Sign In

Communications of the ACM

ACM TechNews

Dataset Bridges Human Vision and Machine Learning


View as: Print Mobile App Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
Artist's conception of computer vision.

A new dataset allows cognitive neuroscientists to better leverage the deep learning models that have significantly improved artificial vision systems.

Credit: alltechasia.com

Researchers at Carnegie Mellon University and Fordham University have created a new dataset that allows cognitive neuroscientists to better leverage the deep learning models that have significantly improved artificial vision systems.

The dataset, called BOLD5000, is made up of brain scans of four volunteers who each viewed 5,000 images taken from two other computer vision datasets: ImageNet and COCO.

While BOLD5000 is much larger than other neuroimaging studies, which often have utilize 100 or fewer images, some researchers think it is still too small.

However, the field's response has been positive so far, as the publicly available BOLD5000 dataset has already been downloaded more than 2,500 times.

A reasonable fMRI dataset would require at least 50,000 stimulus images and many more volunteers to make any progress, since the class of deep neural networks used to analyze visual imagery is trained on millions of images.

From Carnegie Mellon University
View Full Article

 

Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA


 

No entries found