Research map / Detection and segmentation / Open-set and few-shot detection
SAFARI: Adaptive Sequence Transformer for Weakly Supervised Referring Expression Segmentation
Sayan Nag, Koustava Goswami, Srikrishna Karanam
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 gated cross-attention fusion to uni-modal feature extractors (ablation) | SafaRi · RefCOCO val (100% masks) | mIoU: 68.84 -> 72.41 | Improved |
| add attention mask consistency regularization (AMCR) (ablation) | SafaRi · RefCOCO val (10% masks) | mIoU: 60.11 -> 64.02 | Improved |
| add AMCR with full labels (ablation) | SafaRi · RefCOCO val (100% masks) | mIoU: 72.41 -> 73.35 | Improved |
| use AMCR instead of L2 attention loss (ablation) | SafaRi · RefCOCO val (30% masks) | mIoU: 63.48 -> 67.04 | Improved |
| use lightweight X-FACt fusion instead of ALBEF-like fusion layers (ablation) | SafaRi · RefCOCOg val (30% masks) | mIoU: 54.29 -> 55.72 | Improved |
| add Mask Validity Filtering with SpARC to pseudo-labeling (ablation) | SafaRi · RefCOCO val | mIoU: +4.36 mIoU | Improved |
| increase retraining runs with gamma scheduling (ablation) | SafaRi · RefCOCO (30% and 10% labels) | mIoU | Improved |
| add red-box prompting, blur, and spatial reasoning to zero-shot box proposals (SpARC) (ablation) | SafaRi · WSRES | mIoU | Improved |
| increase AMCR weight lambda beyond 0.4 (ablation) | SafaRi · RefCOCO val | mIoU | Hurt |
| train with only 10% masks and boxes versus Partial-RES with 100% boxes | SafaRi · RefCOCO val (10% masks) | mIoU: 64.01 -> 64.02 | No clear effect |
Closest papers by meaning
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.90)
- Boosting Weakly Supervised Referring Image Segmentation via Progressive Comprehension (NeurIPS 2024 · similarity 0.89)
- SAM as the Guide: Mastering Pseudo-Label Refinement in Semi-Supervised Referring Expression Segmentation (ICML 2024 · similarity 0.89)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.89)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.89)
- Hybrid Global-Local Representation with Augmented Spatial Guidance for Zero-Shot Referring Image Segmentation (CVPR 2025 · 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.
