Research map / Detection and segmentation / Open-set and few-shot detection
Dynamic Multimodal Instance Segmentation Guided by Natural Language Queries
Edgar Margffoy-Tuay, Juan C. Perez, Emilio Botero, Pablo Arbelaez
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 skip connections in the upsampling module (ablation) | DMN · UNC testA | mIoU: 49.26 -> 54.83 | Improved |
| add dynamic filter responses F to the multimodal features (ablation) | DMN · UNC testA | mIoU: 50.34 -> 54.83 | Improved |
| add concatenation of r_t in the synthesis module (ablation) | DMN · UNC testA | mIoU: 54.69 -> 54.83 | No clear effect |
| concatenate learned word embedding e_t with h_t (instead of h_t only) (ablation) | DMN · UNC testA | mIoU: 54.80 -> 54.83 | No clear effect |
| concatenate learned word embedding e_t with h_t (instead of h_t only) (ablation) | DMN · UNC testA | Pr@0.9: 4.63 -> 5.12 | Improved |
| add the linguistic query to the visual module (ablation) | DMN · UNC testA | mIoU: 30.92 -> 54.83 | Improved |
| replace LSTMs with SRUs in LM and SM (ablation) | DMN · UNC testA (low resolution) | trainable parameters / training time: 24M -> 9M (LM), 24.2M -> 10M (SM), training time three times lower | Improved |
Closest papers by meaning
- Segmentation from Natural Language Expressions (ECCV 2016 · similarity 0.90)
- Image Segmentation Using Text and Image Prompts (CVPR 2022 · similarity 0.89)
- LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation (NeurIPS 2025 · similarity 0.88)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.88)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.88)
- See Say and Segment: Teaching LMMs to Overcome False Premises (CVPR 2024 · 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.
