Home

Awesome

Predicting Visual Importance Across Graphic Design Types

This repository contains code associated with the project at: http://predimportance.mit.edu/

The model file, in Keras H5 format, can be downloaded here.

The provided Jupyter notebook contains the model specification and shows how to load and run the Keras model on some sample images.

If you use this code, please consider citing:

Fosco, C., Casser, V., Kumar Bedi, A., O'Donovan, P., Hertzmann, A., Bylinskii, Z.
Predicting Visual Importance Across Graphic Design Types. In ACM UIST, 2020.

Questions? Contact camilolu@mit.edu (or alternative contacts)