Salient Object Detection: A Benchmark

Abstract

We extensively compare, qualitatively and quantitatively, 41 state-of-the-art models (29 salient object detection, 10 fixation prediction, 1 objectness, and 1 baseline) over seven challenging data sets for the purpose of benchmarking salient object detection and segmentation methods. From the results obtained so far, our evaluation shows a consistent rapid progress over the last few years in terms of both accuracy and running time. The top contenders in this benchmark significantly outperform the models identified as the best in the previous benchmark conducted three years ago. We find that the models designed specifically for salient object detection generally work better than models in closely related areas, which in turn provides a precise definition and suggests an appropriate treatment of this problem that distinguishes it from other problems. In particular, we analyze the influences of center bias and scene complexity in model performance, which, along with the hard cases for the state-of-the-art models, provide useful hints toward constructing more challenging large-scale data sets and better saliency models. Finally, we propose probable solutions for tackling several open problems, such as evaluation scores and data set bias, which also suggest future research directions in the rapidly growing field of salient object detection.

Probabilistic saliency estimation

“…While supervised approaches, such as those in [45] and [36], have the potential of finding more accurate results, their performance depends on the training process followed and the data that has been exploited for training. Recent works have also indicated that unsupervised saliency detection approaches can compete (or even outperform) supervised methods [43].…”

“…For an extensive survey on the recent state-of-the-art salient object detection algorithms, readers are encouraged to see [37].…”

“…Furthermore, an extensive benchmark comparing the performance of the state-of-the-art in salient object detection is also provided in [43].…”

“…Although =1 is an intuitive selection for equal weighting of precision and recall values, in the salient object detection literature it is common to set = 0.3 in order to emphasize more on precision [37]. The reason behind this choice is the fact that it is relatively easier to achieve high recall than high precision.…”

Correspondence Driven Saliency Transfer

“…Instead of surveying the large volume of literature, which is impractical here, we mainly focus on recent bottom-up saliency methods and top-down models, and 1 http://github.com/shenjianbing/saliencytransfer analyze their properties and limitations. We refer the readers to [23] and [24] for more detailed reviews of the saliency models.…”