Speaker: Shay Perera
Affiliation: Electrical Eng., Technion
This talk addresses the prediction of memorability of a given image.
We study the relation between deep networks for image classification and memorability prediction.
Our study gives rise to three main insights.
First, scene classification plays a bigger role in memorability prediction than object classification.
Second, as object classification CNNs improve, so does image memorability prediction.
Third, it suffices to train a regression layer on top of a CNN for object \& scene recognition to achieve on par results
with those attained by re-training the entire CNN for memorability prediction.
Based on these observations we propose a network that reaches the limit of human performance on the largest existing
dataset for image memorability.
*MSc student under the supervision of Prof. Ayelet Tal and Prof. Lihi Zelnik-Manor