Research map / Video / Video segmentation and tracking
SAM-DAQ: Segment Anything Model with Depth-guided Adaptive Queries for RGB-D Video Salient Object Detection
Jia Lin, Xiaofei Zhou, Jiyuan Liu, Runmin Cong, Guodao Zhang, Zhi Liu, Jiyong 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 |
|---|---|---|---|
| use parallel depth-guided adapters instead of sequential adapters (ablation) | SAM-DAQ PAMIE · RDVS | training memory (G): 91.9 -> 21.0 GB; E-measure 0.860 -> 0.913 | Improved |
| use parallel adapters instead of LoRA (ablation) | SAM-DAQ PAMIE · RDVS | training memory (G): 95.0 -> 21.0 GB; E 0.889 -> 0.913 | Improved |
| add linear depth projector (ablation) | SAM-DAQ · RDVS | E-measure: 0.899 -> 0.913 | Improved |
| add depth modality (multi-modal fusion) (ablation) | SAM-DAQ · RDVS | E-measure: 0.876 -> 0.913 | Improved |
| use sparse-only learnable embeddings instead of dense or both (ablation) | QTM · RDVS | E-measure: 0.875 (dense) / 0.862 (both) -> 0.913 | Improved |
| add iterative video-level query update (ablation) | QTM · RDVS | E-measure: 0.883 -> 0.913 | Improved |
| replace SAM2 memory bank with QTM (ablation) | SAM-DAQ · RDVS | E-measure: 0.853 -> 0.913 | Improved |
| addition update instead of multiply update (ablation) | QTM · RDVS | E-measure: 0.895 -> 0.913 | Improved |
| supervise only highest-level embeddings instead of all three levels (ablation) | SAM-DAQ intermediate supervision · RDVS | E-measure: 0.845 -> 0.913 | Improved |
| increase query hidden dimension beyond 64 (ablation) | QTM · RDVS | E-measure: 0.913 (64) -> 0.874 (128), 0.880 (256) | Hurt |
| increase video-level queries to 10 from 8 (ablation) | QTM · RDVS | E/S/F-measure | Hurt |
| SAM-DAQ vs KAN-SAM (best SAM-based baseline) | SAM-DAQ · RDVS, ViDSOD-100, DViSal | E-measure: 0.888 -> 0.913 (RDVS) | Improved |
Closest papers by meaning
- M4-SAM: Multi-Modal Mixture-of-Experts with Memory-Augmented SAM for RGB-D Video Salient Object Detection (CVPR 2026 · similarity 0.93)
- SAM 2: Segment Anything in Images and Videos (ICLR 2025 · similarity 0.90)
- ViDSOD-100: A New Dataset and a Baseline Model for RGB-D Video Salient Object Detection (IJCV 2024 · similarity 0.90)
- DVSOD: RGB-D Video Salient Object Detection (NeurIPS 2023 · similarity 0.89)
- SAM-I2V: Upgrading SAM to Support Promptable Video Segmentation with Less than 0.2% Training Cost (CVPR 2025 · similarity 0.89)
- Stable Segment Anything Model (ICLR 2025 · similarity 0.87)
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.
