S. Kwong

DCGNet not only overcomes the limitations of traditional SOD methods in underwater settings but also sets a new benchmark in saliency detection performance.

By dynamically adapting convolution kernels based on object region proportions, RDNet achieves state-of-the-art salient object detection in remote sensing images, outperforming existing methods that struggle with scale variations.

By cleverly fusing Fourier and spatial domain information within a diffusion framework, FSCDiff significantly boosts the accuracy of underwater salient object detection, outperforming existing RGB-D methods.