Quality-aware Selective Fusion Network for V-D-T Salient Object Detection

TL;DR

This work tackles VDT salient object detection by addressing quality variation in depth and thermal modalities. It introduces QSF-Net, a three-subnet framework that first extracts multi-scale features, then learns quality-aware region maps via weak supervision, and finally fuses RGB, depth, and thermal cues under region guidance with attention and edge refinement. The approach yields state-of-the-art results on the VDT-2048 dataset across multiple metrics, demonstrating improved robustness in challenging scenes with degraded modality quality. The method advances multi-modal SOD by aligning fusion with modality quality, which is critical for real-world robotics and vision systems that rely on imperfect sensor data.

Abstract

Depth images and thermal images contain the spatial geometry information and surface temperature information, which can act as complementary information for the RGB modality. However, the quality of the depth and thermal images is often unreliable in some challenging scenarios, which will result in the performance degradation of the two-modal based salient object detection (SOD). Meanwhile, some researchers pay attention to the triple-modal SOD task, where they attempt to explore the complementarity of the RGB image, the depth image, and the thermal image. However, existing triple-modal SOD methods fail to perceive the quality of depth maps and thermal images, which leads to performance degradation when dealing with scenes with low-quality depth and thermal images. Therefore, we propose a quality-aware selective fusion network (QSF-Net) to conduct VDT salient object detection, which contains three subnets including the initial feature extraction subnet, the quality-aware region selection subnet, and the region-guided selective fusion subnet. Firstly, except for extracting features, the initial feature extraction subnet can generate a preliminary prediction map from each modality via a shrinkage pyramid architecture. Then, we design the weakly-supervised quality-aware region selection subnet to generate the quality-aware maps. Concretely, we first find the high-quality and low-quality regions by using the preliminary predictions, which further constitute the pseudo label that can be used to train this subnet. Finally, the region-guided selective fusion subnet purifies the initial features under the guidance of the quality-aware maps, and then fuses the triple-modal features and refines the edge details of prediction maps through the intra-modality and inter-modality attention (IIA) module and the edge refinement (ER) module, respectively. Extensive experiments are performed on VDT-2048

Figures (16)

• Figure 1: Examples of VDT-2048 images: the first column shows the visible images, the second column shows the depth images, the third column shows the thermal images, and the last column presents the ground truth (GT).
• Figure 2: Examples of some low-quality images in VDT-2048 dataset: the first column shows the visible images, and the last column presents the ground truth (GT).
• Figure 3: The overall architecture of the proposed QSF-Net, which consists of three components including the initial feature extraction subnet, the quality-aware region selection subnet, and the region-guided selective fusion subnet. Firstly, the initial feature extraction subnet will preliminarily extract the encoder features and generate the initial prediction maps from the three modalities. Secondly, the quality-aware region selection subnet picks out the high-quality and low-quality regions, generating the quality-aware maps in a weakly-supervised way. Finally, the region-guided selective fusion subnet is deployed to purify the multi-modal features and selectively fuse the features under the guidance of the region information in the quality-aware maps.
• Figure 4: The architecture of the initial feature extraction subnet (visible branch).
• Figure 5: The overall architecture of the initial feature extraction subnet.