Research map / Detection and segmentation / Open-set and few-shot detection
DeRIS: Decoupling Perception and Cognition for Enhanced Referring Image Segmentation through Loopback Synergy
Ming Dai, Wenxuan Cheng, Jiang-jiang Liu, Sen Yang, Wenxiao Cai, Yanpeng Sun, Wankou Yang
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 |
|---|---|---|---|
| scale cognition encoder from BERT-B to BEiT3-B (ablation) | DeRIS (Swin-S) · gRefCOCO | cIoU: 56.26 -> 66.31 | Improved |
| scale cognition encoder from BEiT3-B to BEiT3-L (ablation) | DeRIS (Swin-S) · gRefCOCO | cIoU: 66.31 -> 69.14 | Improved |
| scale perception backbone from Swin-T to Swin-B (ablation) | DeRIS (BEiT3-B) · gRefCOCO | cIoU: 65.54 -> 66.78 | No clear effect |
| use Qwen2-7B cognition instead of BEiT3-0.2B (ablation) | DeRIS (Swin-B) · RefCOCO+ | gIoU: 72.43 -> 79.88 | Improved |
| add NSC augmentation (ablation) | DeRIS · gRefCOCO | N-acc: 60.19 -> 75.36 | Improved |
| add non-referent supervision during training (ablation) | DeRIS · gRefCOCO | N-acc: 60.19 -> 62.71 | Improved |
| multiply non-referent score at test time (without NSC) (ablation) | DeRIS · gRefCOCO | N-acc: 62.71 -> 68.80 | Improved |
| multiply non-referent score at test time (with NSC) (ablation) | DeRIS · gRefCOCO | gIoU: 73.77 -> 72.78 | No clear effect |
| use Loopback Synergy instead of P-to-C query transfer (ablation) | DeRIS · gRefCOCO | gIoU: 69.98 -> 71.37 | Improved |
| transfer queries cognition-to-perception instead of P-to-C (ablation) | DeRIS · gRefCOCO | gIoU: 69.98 -> 56.77 | Hurt |
| raise NSC conversion rate Rc from 0% to 15% (ablation) | DeRIS · gRefCOCO | N-acc: 70.25 -> 79.25 | Improved |
| raise NSC conversion rate Rc from 15% to 20% (ablation) | DeRIS · gRefCOCO | cIoU: 67.49 -> 67.07 | Hurt |
Closest papers by meaning
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.88)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.88)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.88)
- SaFiRe: Saccade-Fixation Reiteration with Mamba for Referring Image Segmentation (NeurIPS 2025 · similarity 0.87)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.87)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.86)
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.
