Is Combinational Strategy Better For Image Memorability Prediction

Nowadays, multimedia developed so quickly that we are overwhelmed by all sorts of images in our daily life, including the images on the billboard, internet websites, magazines as well as TV shows. However, some images are easily forgotten while others remain stable memory in people’s mind. As has been proved, images have inner characteristics that decide how memorable they are. So far, many different methods have been proposed to predict the memorability of an image. In this paper, we introduce some combinational strategies to improve the performance for image memorability prediction. It aims to answer whether combinational strategies can obtain better performance for image memorability. Experimental results show that the combinational methods indeed outperform unitary methods with traditional features, but have lower performance than that of CNN-based methods.