Modeling Object Dissimilarity for Deep Saliency Prediction
This addresses the challenge of improving visual attention modeling for computer vision applications, representing an incremental advance by enhancing existing methods with a novel component.
The paper tackles the problem of saliency prediction by modeling object dissimilarity, which previous methods overlooked, and shows that incorporating this into deep networks boosts accuracy, outperforming state-of-the-art models on three benchmarks.
Saliency prediction has made great strides over the past two decades, with current techniques modeling low-level information, such as color, intensity and size contrasts, and high-level ones, such as attention and gaze direction for entire objects. Despite this, these methods fail to account for the dissimilarity between objects, which affects human visual attention. In this paper, we introduce a detection-guided saliency prediction network that explicitly models the differences between multiple objects, such as their appearance and size dissimilarities. Our approach allows us to fuse our object dissimilarities with features extracted by any deep saliency prediction network. As evidenced by our experiments, this consistently boosts the accuracy of the baseline networks, enabling us to outperform the state-of-the-art models on three saliency benchmarks, namely SALICON, MIT300 and CAT2000. Our project page is at https://github.com/IVRL/DisSal.