Salient Object Detection in the Deep Learning Era: An In-Depth Survey

• DOI: 10.48550/arxiv.1904.09146
• Publication Date: 2019-01-01
• Authors: Wenguan Wang; Qiuxia Lai; Huazhu Fu; Jianbing Shen; Haibin Ling; Ruigang Yang
• Keywords: Robustness; Benchmark (surveying)

As an essential problem in computer vision, salient object detection (SOD) has attracted increasing amount of research attention over the years. Recent advances SOD are predominantly led by deep learning-based solutions (named SOD). To enable in-depth understanding SOD, this paper, we provide a comprehensive survey covering various aspects, ranging from algorithm taxonomy to unsolved issues. In particular, first review algorithms different perspectives, including network architecture, level supervision, learning paradigm, and object-/instance-level detection. Following that, summarize analyze existing datasets evaluation metrics. Then, benchmark large group representative models, detailed analyses comparison results. Moreover, study performance under attribute settings, which not been thoroughly explored previously, constructing novel dataset with rich annotations types, challenging factors, scene categories. We further analyze, for time field, robustness models random input perturbations adversarial attacks. also look into generalization difficulty datasets. Finally, discuss several open issues outline future directions.