Research map / Vision-language and multimodal / Cross-modal learning
Cross-Modal Self-Attention Network for Referring Image Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 cross-modal self-attention over multimodal features instead of no attention (ablation) | CMSA multimodal representation · UNC val | IoU: 45.63 -> 50.12 | Helped |
| use cross-modal self-attention instead of word attention (ablation) | CMSA multimodal representation · UNC val | IoU: 47.01 -> 50.12 | Helped |
| use cross-modal self-attention instead of pixel attention (ablation) | CMSA multimodal representation · UNC val | IoU: 47.84 -> 50.12 | Helped |
| use cross-modal self-attention instead of word-pixel pair attention (ablation) | CMSA multimodal representation · UNC val | IoU: 47.57 -> 50.12 | Helped |
| retain word-level language features rather than encoding the whole expression as one LSTM vector (ablation) | CMSA · UNC val | IoU: 48.53 -> 50.12 | Helped |
| use gated multi-level fusion instead of PPM (ablation) | CMSA · UNC val | IoU: 53.54 -> 58.32 | Helped |
| use gated multi-level fusion instead of Deconv (ablation) | CMSA · UNC val | IoU: 54.18 -> 58.32 | Helped |
| use gated multi-level fusion instead of ConvLSTM (ablation) | CMSA · UNC val | IoU: 56.56 -> 58.32 | Helped |
| allow gated interactions between feature levels rather than self-gating each level independently (ablation) | CMSA multi-level fusion · UNC val | IoU: 57.08 -> 58.32 | Helped |
| use CMSA word-level representation instead of the RMI-LSTM representation (ablation) | multimodal feature representation · UNC val | IoU: 45.18 -> 50.12 | Helped |
| use CMSA word-level representation instead of the RRN-CNN representation (ablation) | multimodal feature representation · UNC val | IoU: 46.95 -> 50.12 | Helped |
Closest papers by meaning
- X-RefSeg3D: Enhancing Referring 3D Instance Segmentation via Structured Cross-Modal Graph Neural Networks (AAAI 2024 · similarity 0.88)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.88)
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.87)
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.86)
- COSMOS: Cross-Modality Self-Distillation for Vision Language Pre-training (CVPR 2025 · similarity 0.86)
- Prompting Multi-Modal Image Segmentation with Semantic Grouping (AAAI 2024 · 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.