DNA: Deeply-supervised Nonlinear Aggregation for Salient Object Detection

TL;DR

This paper introduces Deeply-supervised Nonlinear Aggregation (DNA), a novel method that improves salient object detection by effectively integrating multi-scale features through nonlinear aggregation, outperforming existing linear methods.

Contribution

The paper proposes a new nonlinear aggregation approach for deep supervision in CNNs, enhancing the utilization of multi-scale features for salient object detection.

Findings

DNA outperforms state-of-the-art methods on multiple datasets.

Nonlinear aggregation better leverages side-output information.

The modified U-Net with DNA achieves superior accuracy.

Abstract

Recent progress on salient object detection mainly aims at exploiting how to effectively integrate multi-scale convolutional features in convolutional neural networks (CNNs). Many popular methods impose deep supervision to perform side-output predictions that are linearly aggregated for final saliency prediction. In this paper, we theoretically and experimentally demonstrate that linear aggregation of side-output predictions is suboptimal, and it only makes limited use of the side-output information obtained by deep supervision. To solve this problem, we propose Deeply-supervised Nonlinear Aggregation (DNA) for better leveraging the complementary information of various side-outputs. Compared with existing methods, it i) aggregates side-output features rather than predictions, and ii) adopts nonlinear instead of linear transformations. Experiments demonstrate that DNA can successfully break through the bottleneck of current linear approaches. Specifically, the proposed saliency detector, a modified U-Net architecture with DNA, performs favorably against state-of-the-art methods on various datasets and evaluation metrics without bells and whistles.