.png){kind=icon}
Return
Survey on deep learning-based weakly supervised salient object detection
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.127497.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Salient Object Detection (SOD) in images (ISOD) and videos (VSOD) has always been an active field of research in computer vision. While fully supervised SOD algorithms have made significant progress over the years, they heavily rely on manual pixel-level annotations, which are time-consuming and labor-intensive. Weakly supervised learning paradigms have been explored in both image-based and video-based scenarios, leading to Weakly Supervised ISOD (WSISOD) and Weakly Supervised VSOD (WSVSOD) respectively. These algorithms aim to strike a balance between annotation cost and model performance by leveraging low-cost weak annotations as supervision. Based on the systematic analysis of research challenges in WSISOD and WSVSOD algorithms, this paper presents a comprehensive survey of recent advances in both deep learning-based WSISOD and WSVSOD algorithms. Furthermore, currently available benchmark ISOD and VSOD datasets were summarized, the evaluation metrics of the performance and the performance of existing ISOD and VSOD algorithms on different datasets were depicted. Finally, the existing problems and the future tendency of weakly supervised salient object detection research were discussed.
Keywords:
WSISOD
WSVSOD
Deep learning
Weak annotations
Journal
IF:
7.5
Papers:
3.0W
Citations:
10.2W
Organization
Cited Papers
.png){kind=icon}
ShareSaveWeakly Supervised Instance Segmentation of Electrical Equipment Based on RGB-T Automatic Annotation
PREAI
ShareSaveShareSaveShareSave
ShareSave
View more