Multiscale Deep Network with Centerness-Aware Loss for Salient Object Detection

<div>Illustration of the detailed FCL (fully connected layer). The first column contains 8192 nodes, the second column contains 100 nodes, and the third column contains only one node.</div>

Advances in Multimedia

fig6

Figure 6

Figure 6: Multiscale Deep Network with Centerness-Aware Loss for Salient Object Detection 