Research map / Detection and segmentation / CNN features
A Single Stream Network for Robust and Real-time RGB-D Salient Object Detection
Xiaoqi Zhao, Lihe Zhang¹, Youwei Pang, Huchuan Lu, Lei Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add 4-channel RGB-D concatenation with He-initialized first layer (Cat_He) (ablation) | FPN VGG-16 baseline · RGBD135 | F_beta^max: +4.53% | Improved |
| add 4-channel RGB-D concatenation with He-initialized first layer (Cat_He) (ablation) | FPN VGG-16 baseline · RGBD135 | F_beta^mean: +5.44% | Improved |
| add 4-channel RGB-D concatenation with He-initialized first layer (Cat_He) (ablation) | FPN VGG-16 baseline · RGBD135 | F_beta^w: +5.25% | Improved |
| add 4-channel RGB-D concatenation with He-initialized first layer (Cat_He) (ablation) | FPN VGG-16 baseline · RGBD135 | MAE: 16.36% | Improved |
| use concatenation (Cat_He) instead of element-wise addition early fusion (ablation) | early fusion · five RGB-D datasets | F-measure, S-measure, MAE | Improved |
| use raw depth map directly as attention between encoder and decoder (ablation) | Cat_He model · five RGB-D datasets | saliency metrics | Hurt |
| use mask-guided attention instead of depth as attention (ablation) | Cat_He model · five RGB-D datasets | saliency metrics | Improved |
| add depth guidance with dual attended branches (DEDA) on top of mask-guided attention (ablation) | MGA · five RGB-D datasets | saliency metrics | Improved |
| add background branch (DEDA vs DEFA) (ablation) | DEFA · five RGB-D datasets | saliency metrics | Improved |
| replace ASPP with PAFE (ablation) | ASPP · five RGB-D datasets | saliency metrics: FLOPs 3.86G -> 4.00G; Params 6.82M -> 7.07M | Improved |
| use single-stream encoder with VGG-16 instead of two-stream models | RGB-D SOD · model size | parameters: saves 55.5% of the parameters compared to DMRA | Improved |
Closest papers by meaning
- A2dele: Adaptive and Attentive Depth Distiller for Efficient RGB-D Salient Object Detection (CVPR 2020 · similarity 0.94)
- RGB-D Salient Object Detection via 3D Convolutional Neural Networks (AAAI 2021 · similarity 0.94)
- Asymmetric Two-Stream Architecture for Accurate RGB-D Saliency Detection (ECCV 2020 · similarity 0.92)
- Calibrated RGB-D Salient Object Detection (CVPR 2021 · similarity 0.91)
- Deep RGB-D Saliency Detection With Depth-Sensitive Attention and Automatic Multi-Modal Fusion (CVPR 2021 · similarity 0.91)
- Accurate RGB-D Salient Object Detection via Collaborative Learning (ECCV 2020 · similarity 0.91)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
