CMGNet: Context-aware middle-layer guidance network for salient object detection
Salient object detection (SOD) is a critical task in computer vision that involves accurately identifying and segmenting visually significant objects in an image. To address the challenges of gridding issues and feature dilution effects commonly encountered in SOD, we propose a sophisticated context...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-01-01
|
| Series: | Journal of King Saud University: Computer and Information Sciences |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1319157823003920 |
| _version_ | 1827356751984328704 |
|---|---|
| author | Inam Ullah Sumaira Hussain Kashif Shaheed Wajid Ali Shahid Ali Khan Yilong Yin Yuling Ma |
| author_facet | Inam Ullah Sumaira Hussain Kashif Shaheed Wajid Ali Shahid Ali Khan Yilong Yin Yuling Ma |
| author_sort | Inam Ullah |
| collection | DOAJ |
| description | Salient object detection (SOD) is a critical task in computer vision that involves accurately identifying and segmenting visually significant objects in an image. To address the challenges of gridding issues and feature dilution effects commonly encountered in SOD, we propose a sophisticated context-aware middle-layer guidance network (CMGNet). CMGNet incorporates the context-aware central-layer guidance module (CCGM), which utilizes cost-effective large kernels of depth-wise convolutions with embedded parallel channel attentions and squeeze-and-excitation (SeE) attentions mechanisms. It enables the model to effectively perceive objects of varying scales in complex scenarios. Additionally, the incorporation of the adjacent-to-central-layers paradigm enriches the model’s ability to capture more structural and contextual information. To further enhance performance, we introduce the dual-phase central-layer refinement module (DCRM), which effectively removes the minute blurry residuals in complex scenarios and enhances object segmentation. Moreover, we propose a novel hybrid loss function that handles hard pixels at or near boundaries by incorporating a weighting formula. This hybrid loss function combines binary cross-entropy (BCE), intersection over union (IoU), and consistency-enhanced loss (CEL), resulting in smoother and more precise saliency maps. Extensive evaluations on challenging datasets demonstrate the superiority of our approach over 15 state-of-the-art methods in salient object detection. |
| first_indexed | 2024-03-08T05:14:30Z |
| format | Article |
| id | doaj.art-ca66b19c50d144f2aa54a8e469b21e97 |
| institution | Directory Open Access Journal |
| issn | 1319-1578 |
| language | English |
| last_indexed | 2024-03-08T05:14:30Z |
| publishDate | 2024-01-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Journal of King Saud University: Computer and Information Sciences |
| spelling | doaj.art-ca66b19c50d144f2aa54a8e469b21e972024-02-07T04:42:44ZengElsevierJournal of King Saud University: Computer and Information Sciences1319-15782024-01-01361101838CMGNet: Context-aware middle-layer guidance network for salient object detectionInam Ullah0Sumaira Hussain1Kashif Shaheed2Wajid Ali3Shahid Ali Khan4Yilong Yin5Yuling Ma6School of Computer Science and Technology, Shandong Jianzhu University, Jinan 250101, ChinaSchool of Computer Science and Technology, Shandong Jianzhu University, Jinan 250101, ChinaDepartment of Multimedia Systems, Faculty of Electronics, Telecommunication and Informatics, Gdansk University of Technology, 80-233 Gdansk, PolandCollege of Computer Science and Technology, Jilin University, Changchun 130012, Jilin, ChinaDepartment of Computer Science, Quaid-e-Azam University, Islamabad, PakistanSchool of Computer Science and Technology, Shandong Jianzhu University, Jinan 250101, China; School of Software, Shandong University, Jinan 250101, ChinaSchool of Computer Science and Technology, Shandong Jianzhu University, Jinan 250101, China; Corresponding author.Salient object detection (SOD) is a critical task in computer vision that involves accurately identifying and segmenting visually significant objects in an image. To address the challenges of gridding issues and feature dilution effects commonly encountered in SOD, we propose a sophisticated context-aware middle-layer guidance network (CMGNet). CMGNet incorporates the context-aware central-layer guidance module (CCGM), which utilizes cost-effective large kernels of depth-wise convolutions with embedded parallel channel attentions and squeeze-and-excitation (SeE) attentions mechanisms. It enables the model to effectively perceive objects of varying scales in complex scenarios. Additionally, the incorporation of the adjacent-to-central-layers paradigm enriches the model’s ability to capture more structural and contextual information. To further enhance performance, we introduce the dual-phase central-layer refinement module (DCRM), which effectively removes the minute blurry residuals in complex scenarios and enhances object segmentation. Moreover, we propose a novel hybrid loss function that handles hard pixels at or near boundaries by incorporating a weighting formula. This hybrid loss function combines binary cross-entropy (BCE), intersection over union (IoU), and consistency-enhanced loss (CEL), resulting in smoother and more precise saliency maps. Extensive evaluations on challenging datasets demonstrate the superiority of our approach over 15 state-of-the-art methods in salient object detection.http://www.sciencedirect.com/science/article/pii/S1319157823003920Salient object detectionSOD tasksLightweight salient object detectionMulti-scale learningDeep learning |
| spellingShingle | Inam Ullah Sumaira Hussain Kashif Shaheed Wajid Ali Shahid Ali Khan Yilong Yin Yuling Ma CMGNet: Context-aware middle-layer guidance network for salient object detection Journal of King Saud University: Computer and Information Sciences Salient object detection SOD tasks Lightweight salient object detection Multi-scale learning Deep learning |
| title | CMGNet: Context-aware middle-layer guidance network for salient object detection |
| title_full | CMGNet: Context-aware middle-layer guidance network for salient object detection |
| title_fullStr | CMGNet: Context-aware middle-layer guidance network for salient object detection |
| title_full_unstemmed | CMGNet: Context-aware middle-layer guidance network for salient object detection |
| title_short | CMGNet: Context-aware middle-layer guidance network for salient object detection |
| title_sort | cmgnet context aware middle layer guidance network for salient object detection |
| topic | Salient object detection SOD tasks Lightweight salient object detection Multi-scale learning Deep learning |
| url | http://www.sciencedirect.com/science/article/pii/S1319157823003920 |
| work_keys_str_mv | AT inamullah cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT sumairahussain cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT kashifshaheed cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT wajidali cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT shahidalikhan cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT yilongyin cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection AT yulingma cmgnetcontextawaremiddlelayerguidancenetworkforsalientobjectdetection |