Research map / Vision-language and multimodal / Human and embodied modeling
Areas of Attention for Image Captioning
Marco Pedersoli, Thomas Lucas, Cordelia Schmid, Jakob Verbeek
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 word-region interaction term (ablation) | captioning baseline, activation grid · MSCOCO | CIDEr: 78.9 -> 83.6 | Improved |
| add region-state interaction term (ablation) | attention model, activation grid · MSCOCO | CIDEr: 83.6 -> 85.5 | Improved |
| add visual feedback in the RNN state update (ablation) | attention model, activation grid · MSCOCO | CIDEr: 85.5 -> 87.4 | Improved |
| use word-conditional feedback instead of marginal feedback (ablation) | attention model, activation grid · MSCOCO | CIDEr: 87.4 -> 86.8 | Hurt |
| replace activation grid with spatial transformer regions (RNN training only) | attention model · MSCOCO | CIDEr: 87.4 -> 91.1 | Improved |
| replace object proposals with spatial transformer regions (RNN training only) | attention model · MSCOCO | CIDEr: 89.0 -> 91.1 | Improved |
| replace activation grid with object proposals (RNN training only) | attention model · MSCOCO | CIDEr: 87.4 -> 89.0 | Improved |
| add joint CNN-RNN fine-tuning (ablation) | attention model, spatial transformers · MSCOCO | CIDEr: 91.1 -> 93.8 | Improved |
| add joint CNN-RNN fine-tuning (ablation) | baseline captioning model · MSCOCO | CIDEr: 78.9 -> 87.1 | Improved |
| replace object proposals with spatial transformers after fine-tuning | attention model · MSCOCO | CIDEr: 93.7 -> 93.8 | No clear effect |
| replace activation grid with spatial transformers after fine-tuning | attention model · MSCOCO | CIDEr: 92.6 -> 93.8 | Improved |
| use four or fewer test-time regions | attention model · MSCOCO | CIDEr | Hurt |
Closest papers by meaning
- Attention-Aligned Transformer for Image Captioning (AAAI 2022 · similarity 0.91)
- Area Attention (ICML 2019 · similarity 0.91)
- Text-Guided Attention Model for Image Captioning (AAAI 2017 · similarity 0.90)
- Normalized and Geometry-Aware Self-Attention Network for Image Captioning (CVPR 2020 · similarity 0.90)
- Image Captioning: Transforming Objects into Words (NeurIPS 2019 · similarity 0.90)
- Boosted Attention: Leveraging Human Attention for Image Captioning (ECCV 2018 · similarity 0.90)
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.
