Research map / Vision-language and multimodal / Human and embodied modeling
Robust Change Captioning
Dong Huk Park, Trevor Darrell, Anna Rohrbach
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 |
|---|---|---|---|
| use representation difference instead of pixel difference | Capt-Pix-Diff · CLEVR-Change (total) | CIDEr: 75.9 -> 87.9 | Improved |
| add a single spatial attention to the captioner (ablation) | Capt-Rep-Diff · CLEVR-Change (total) | CIDEr: 87.9 -> 106.4 | Improved |
| use two separate spatial attentions instead of one (ablation) | Capt-Att · CLEVR-Change (total) | CIDEr: 106.4 -> 108.5 | Improved |
| use two separate spatial attentions instead of one (ablation) | Capt-Att · CLEVR-Change (pointing game, total) | Pointing game accuracy (%): 39.37 -> 39.35 | No clear effect |
| add the Dynamic Speaker on top of Dual Attention (ablation) | Capt-Dual-Att · CLEVR-Change (total) | CIDEr: 108.5 -> 112.3 | Improved |
| add the Dynamic Speaker on top of Dual Attention (ablation) | Capt-Dual-Att · CLEVR-Change (pointing game, total) | Pointing game accuracy (%): 39.35 -> 48.10 | Improved |
| add the Dynamic Speaker on top of Dual Attention (ablation) | Capt-Dual-Att · CLEVR-Change (scene change) | CIDEr: 89.8 -> 94.6 | Improved |
| add the Dynamic Speaker on top of Dual Attention (ablation) | Capt-Dual-Att · CLEVR-Change (distractor) | CIDEr: 108.9 -> 110.8 | Improved |
| add a single spatial attention to the captioner (ablation) | Capt-Rep-Diff · CLEVR-Change (scene change) | CIDEr: 51.8 -> 87.2 | Improved |
| add a single spatial attention to the captioner (ablation) | Capt-Rep-Diff · CLEVR-Change (distractor) | CIDEr: 105.3 -> 106.6 | Improved |
| use DUDA with Dual Attention on precomputed clusters instead of DDLA | DDLA · Spot-the-Diff (single change) | CIDEr: 0.340 -> 0.325 | Hurt |
| use DUDA with Dual Attention on precomputed clusters instead of DDLA | DDLA · Spot-the-Diff (single change) | ROUGE-L: 0.283 -> 0.291 | Improved |
Closest papers by meaning
- Region-aware Difference Distilling with Attribute-guided Contrastive Regularization for Change Captioning (AAAI 2025 · similarity 0.88)
- VisMin: Visual Minimal-Change Understanding (NeurIPS 2024 · similarity 0.87)
- DECIDER: Difference-aware Contrastive Diffusion Model with Adversarial Perturbations for Image Change Captioning (AAAI 2025 · similarity 0.86)
- Revisiting Change Captioning from Self-supervised Global-Part Alignment (AAAI 2025 · similarity 0.86)
- Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective (CVPR 2025 · similarity 0.85)
- Describing Differences in Image Sets with Natural Language (CVPR 2024 · similarity 0.85)
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.
