Antonio Torralba

TL;DR: Motion magnification as discussed by the authors is a technique that acts like a microscope for visual motion, which can amplify subtle motions in a video sequence, allowing for visualization of deformations that would otherwise be invisible.

TL;DR: A database of 360° panoramic images organized into 26 place categories is constructed, and the problem of scene viewpoint recognition is introduced, to classify the type of place shown in a photo, and also recognize the observer's viewpoint within that category of place.

TL;DR: In this article, the authors identify a group of interpretable units that are closely related to object concepts using a segmentation-based network dissection method and quantitatively quantify the causal effect of these units by measuring the ability of interventions to control objects in the output.

TL;DR: This work puts forth a benchmark for computational models of search in real world scenes by recording observers’ eye movements as they performed a search task (person detection) in 912 outdoor scenes and finding that observers were highly consistent in the regions fixated during search.

TL;DR: LaMem is built, the largest annotated image memorability dataset to date, using Convolutional Neural Networks, to demonstrate that one can now robustly estimate the memorability of images from many different classes, positioning memorability and deep memorability features as prime candidates to estimate the utility of information for cognitive systems.