Viraliency: Pooling Local Virality

Abstract

In our overly-connected world, the automatic recognition of virality -- the quality of an image or video to be rapidly and widely spread -- is of crucial importance, and has recently awaken the interest of the computer vision community Concurrently, recent progress in deep learning architectures showed that global (average) pooling strategies allow to extract class activation maps, which highlight the part of the image most likely to contain a certain class. We extend this concept by introducing a pooling layer that learns the size of the average support: the learned top-N average (LENA) pooling. We hypothesize that the latent concepts (feature maps) describing virality may require such a rich pooling strategy and perform an extensive evaluation to assess the validity of this hypothesis. Moreover, we also appraise the use of objectness maps at predicting and localizing the virality of an image. Experiments are shown in two publicly available datasets annotated for virality.