Salient Object Detection Based on Deep Residual Networks and Edge Supervised Learning

Feifei Shi; Songlong Zhang; Li Peng

doi:10.3788/LOP56.151502

Laser & Optoelectronics Progress, Volume. 56, Issue 15, 151502(2019)

Salient Object Detection Based on Deep Residual Networks and Edge Supervised Learning

Feifei Shi^**, Songlong Zhang, and Li Peng^*

Engineering Research Center of Internet of Things Technology Applications of the Ministry of Education, College of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(9)

Fig. 1. Network structure

Download full size

Fig. 2. Edge residuals block structure diagram

Download full size

Fig. 3. Multi-scale atrous convolution unit

Download full size

Fig. 4. Contrast of ground truth image. (a) Original ground truth image; (b) modified manual marking diagram

Download full size

Fig. 5. Contrast of three-category model proposed in our paper and the traditional two-category model. (a) Original image; (b) result of traditional two-classification model; (c) results of proposed three-classification model; (d) ground truth image

Download full size

Fig. 6. Comparison of saliency maps. (a) Original image; (b) BL; (c) BSCA; (d) RFCN;(e) MDF; (f) DCL; (g) DHS; (h) UCF; (i) proposed method; (j) ground truth image

Download full size

Fig. 7. Precision-recall curves of the proposed algorithm and other state-of-the-art methods. (a) Results on ECSSD dataset; (b) results on SED2 dataset

Download full size

Table 1. Performance comparison of proposed modules
View table
Table 1. Performance comparison of proposed modules
Module F-measure MAE
ResNet-101+2 classification 0.7747 0.0856
ResNet-101+ERB 0.8537 0.0715
ResNet-101+3 classification 0.8715 0.0774
ResNet-101+ERB+3 classification 0.9048 0.0595

Table 2. Performance comparison of each algorithm
View table
Table 2. Performance comparison of each algorithm
Item BL BSCA RFCN MDF DCL DHS UCF Proposed
MAE 0.216 0.182 0.107 0.105 0.074 0.060 0.078 0.059
F-measure 0.760 0.753 0.866 0.869 0.901 0.889 0.910 0.905
AUC 0.916 0.922 0.976 0.947 0.971 0.972 0.980 0.981

Tools

Get Citation

Copy Citation Text

Feifei Shi, Songlong Zhang, Li Peng. Salient Object Detection Based on Deep Residual Networks and Edge Supervised Learning[J]. Laser & Optoelectronics Progress, 2019, 56(15): 151502

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites