Search for: All records

This paper presents an approach for answering fill-in-the-blank multiple choice questions from the Visual Madlibs dataset.Instead of generic and commonly used representations trained on the ImageNet classification task, our approach employs acombination of networks trained for specialized tasks such as scene recognition, person activity classification, and attributeprediction. We also present a method for localizing phrases from candidate answers in order to provide spatial support forfeature extraction. We map each of these features, together with candidate answers, to a joint embedding space throughnormalized canonical correlation analysis (nCCA). Finally, we solve an optimization problem to learn to combine scoresfrom nCCA models trained on multiple cues to select the best answer. Extensive experimental results show a significantimprovement over the previous state of the art and confirm that answering questions from a wide range of types benefits fromexamining a variety of image cues and carefully choosing the spatial support for feature extraction.