Research map / Detection and segmentation / Open-set and few-shot detection
AMLRIS: Alignment-aware Masked Learning for Referring Image Segmentation
Tongfei Chen, Shuo Yang, Yuguang Yang, Linlin Yang, Runtang Guo, Changbai Li, He Long, Chunyu Xie, Dawei Leng, Baochang 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 alignment-aware masked learning (AML) | CARIS · RefCOCO+ val | oIoU: 65.54 -> 67.37 | Improved |
| add AML | DETRIS · RefCOCO+ val | oIoU: 65.39 -> 66.20 | Improved |
| add AML | ReLA · GRES early stage | cIoU: 28.49 -> 30.86 | Improved |
| compute PMME similarity early (before fusion) rather than late (ablation) | AML · RefCOCO+ val | oIoU: 63.35 -> 67.37 | Improved |
| add Late PMME to CARIS (ablation) | CARIS · RefCOCO+ val | oIoU: 65.54 -> 63.35 | Hurt |
| use random projection rather than learnable projection in PMME (ablation) | AML · RefCOCO+ epoch 1 | oIoU: 16.92 -> 24.85 | Improved |
| use learnable projection instead of CARIS baseline (ablation) | CARIS · RefCOCO+ epoch 1 | oIoU: 20.24 -> 16.92 | Hurt |
| raise threshold tau from 0 to 0.4 (ablation) | AML · RefCOCO+ val | oIoU: 65.54 -> 67.37 | Improved |
| raise threshold tau from 0.3 to 0.4 (ablation) | AML · RefCOCO val | oIoU: 75.81 -> 75.45 | Hurt |
| increase projection dim D_a from 1024 to 2048 (ablation) | AML · RefCOCO+ val | oIoU: 66.82 -> 67.37 | Improved |
| increase D_a from 2048 to 4096 (ablation) | AML · RefCOCO+ val | oIoU: 67.37 -> 67.08 | No clear effect |
| train with AML masking, test under occlusion | CARIS · RefCOCO testA occlusion | oIoU drop: 9.4% -> 3.7% | Improved |
Closest papers by meaning
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.91)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · similarity 0.89)
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.89)
- OneRef: Unified One-tower Expression Grounding and Segmentation with Mask Referring Modeling (NeurIPS 2024 · similarity 0.88)
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.
