Research map / Image generation / Editing and style transfer
Generating More Pertinent Captions by Leveraging Semantics and Style on Multi-Source Datasets
Marcella Cornia, Lorenzo Baraldi, Giuseppe Fiameni, Rita Cucchiara
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 web-collected training data (Tiny) (ablation) | Ours Tiny · COCO test | CIDEr: 139.4 -> 148.4 | Improved |
| add web-collected training data (Base) (ablation) | Ours Base · nocaps validation overall | CIDEr: 89.2 -> 120.5 | Improved |
| add retrieval keywords alone (Tiny, no style token) (ablation) | Ours Tiny · COCO test | CIDEr: 145.8 -> 146.5 | Improved |
| add human-annotated style token on top of keywords (Tiny) (ablation) | Ours Tiny · COCO test | CIDEr: 146.5 -> 148.4 | Improved |
| add keywords and style token (Base) (ablation) | Ours Base · nocaps validation overall | CIDEr: 114.7 -> 120.5 | Improved |
| add keywords and style token (Base) (ablation) | Ours Base · COCO test | CIDEr: 147.9 -> 149.6 | Improved |
| retrieval keywords instead of Faster R-CNN tags (Tiny) (ablation) | Ours · COCO test (COCO only) | CIDEr: 141.1 -> 141.9 | Improved |
| retrieval keywords instead of Faster R-CNN tags (Base) (ablation) | Ours Base · COCO test (COCO only) | CIDEr: 140.5 -> 143.4 | Improved |
| condition on human-annotated rather than web-collected style token | Ours Base · COCO test | CIDEr: 86.3 -> 132.3 | Improved |
| condition on web-collected rather than human-annotated style token | Ours Base · CC3M validation | CIDEr: 58.8 -> 105.7 | Improved |
| scale training images from 5.8M to 35.7M | Ours Base · COCO test | CIDEr: 147.8 -> 149.6 | Improved |
| scale training images from 5.8M to 35.7M | Ours Base · nocaps validation overall | CIDEr: 118.0 -> 120.5 | Improved |
Closest papers by meaning
- Generating Diverse and Descriptive Image Captions Using Visual Paraphrases (ICCV 2019 · similarity 0.90)
- A Picture is Worth More Than 77 Text Tokens: Evaluating CLIP-Style Models on Dense Captions (CVPR 2024 · similarity 0.88)
- CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and Thoroughness (NeurIPS 2025 · similarity 0.87)
- Enhancing Descriptive Captions with Visual Attributes for Multimodal Perception (CVPR 2026 · similarity 0.87)
- Image Textualization: An Automatic Framework for Generating Rich and Detailed Image Descriptions (NeurIPS 2024 · similarity 0.87)
- BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs (CVPR 2025 · 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.
