Research map / Detection and segmentation / Open-set and few-shot detection
Object Relation Attention for Image Paragraph Captioning
Read the paperCheck 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 |
|---|---|---|---|
| Add asymmetric relation-aware features for overlapping object pairs | Top-Down Attention LSTM paragraph captioner · Stanford paragraph test | CIDEr: 30.63 -> 33.38 | Helped |
| Use only overlapping object pairs rather than all pairs (ablation) | Asymmetric relation network · Stanford paragraph test | CIDEr: 30.88 -> 33.12 | Helped |
| Use asymmetric pair features rather than symmetric pair features (ablation) | Relation network with overlapping pairs · Stanford paragraph test | CIDEr: 32.98 -> 33.38 | Helped |
| Use asymmetric splitting rather than asymmetric-to-symmetric encoding (ablation) | Relation network with overlapping pairs · Stanford paragraph test | CIDEr: 32.20 -> 33.38 | Helped |
| Add relation encoding and relation attention (ablation) | SCST paragraph captioner with repetition penalty · Stanford paragraph test | CIDEr: 30.63 -> 33.38 | Helped |
| Add relation encoding when relation attention is disabled (ablation) | SCST paragraph captioner with repetition penalty · Stanford paragraph test | CIDEr: 30.63 -> 31.80 | Helped |
| Add relation attention to the relation-encoding configuration (ablation) | SCST paragraph captioner with repetition penalty · Stanford paragraph test | CIDEr: 31.80 -> 33.38 | Helped |
| Add repetition penalty to cross-entropy training (ablation) | Paragraph captioner · Stanford paragraph test | CIDEr: 12.89 -> 16.27 | Helped |
| Add relation encoding and attention to repetition-penalty cross-entropy training (ablation) | Paragraph captioner · Stanford paragraph test | CIDEr: 16.27 -> 22.85 | Helped |
| Replace raw object features with relation-aware features | Anderson et al. image captioner · MS COCO c5 test | CIDEr: 117.9 -> 122.0 | Helped |
Closest papers by meaning
- A Hierarchical Approach for Generating Descriptive Image Paragraphs (CVPR 2017 · similarity 0.90)
- Image Captioning: Transforming Objects into Words (NeurIPS 2019 · similarity 0.88)
- Learning Object Context for Dense Captioning (AAAI 2019 · similarity 0.87)
- Deep Compositional Captioning: Describing Novel Object Categories Without Paired Training Data (CVPR 2016 · similarity 0.87)
- Relational Distant Supervision for Image Captioning without Image-Text Pairs (AAAI 2024 · similarity 0.87)
- Composing Object Relations and Attributes for Image-Text Matching (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.