This is ResMem: A state-of-the-art machine learning model for predicting the intrinsic memorability of an image. Simply upload an image below and it will be sent to a server, and the estimated memorability of your image will be sent back in a few moments. More information on ResMem and its development can be found here. You can also read our preprint here.

This web app is designed as a demonstration. If you’d like to use ResMem to process a large quantity of images, please use the resmem package for python. The package is available here, and can be downloaded with

pip install resmem

If you use ResMem for your research, please cite our preprint (there is also a sample citation for the package at the bottom of the page):

Needell, C. D., & Bainbridge, W. A. (2021). Embracing New Techniques in Deep Learning for Estimating Image Memorability. ArXiv:2105.10598 [Cs].
We would love to post a link / citation to your results if you have tried ResMem with an alternate stimulus set — please e-mail us if you have! We hope to have a list of all successful (and unsuccessful) validations of the model, which will serve as a useful resource to the research community.



@software{Needell_2021, title={resmem: A package that wraps the ResMem pytorch model.}, url={}, author={Needell, Coen D.}, year={2021}, month={May} }