Understanding the Intrinsic Memorability of Images.
Isola, P., Parikh, D., Torralba, A., & Oliva, A. (2011). NIPS
Artists, advertisers and photographers are routinely challenged by the question ``what makes an image memorable?'' and are then presented with the task of how to create an image that speaks to the viewer. While it may seem like image memorability is purely subjective, recent work shows that it is not an inexplicable phenomenon: variation in memorability of images is consistent across subjects, suggesting that some images are intrinsically more memorable than others, independent of a subject's context and biases. In this paper, we used the publicly available memorability dataset of Isola et al (2011), and augment the object and scene level annotations with interpretable attributes. After identifying a collection of attributes that describe various spatial, content and aesthetic properties of images, we used a feature-selection scheme with desirable explaining-away properties to determine a compact set of features that characterize the memorability of any individual image. We find that images of enclosed spaces containing people with visible-faces are memorable, while images of vistas and peaceful looking places are not. Contrary to popular belief, unusual or aesthetically pleasing scenes are not highly memorable. This work represents one of the first attempts at determining intrinsic image memorability, and opens a new domain of investigation at the interface between human cognition and computer vision.