Research map / Vision-language and multimodal / Human and embodied modeling
Event-Specific Image Importance
Check 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 |
|---|---|---|---|
| use the proposed progressive piecewise-ranking CNN instead of the strongest non-oracle baseline | event-specific image importance prediction · CUFED test set | MAP@5%: 0.266 -> 0.302 | Helped |
| add the initial shared-output training stage before event-specific training (ablation) | PR-CNN · CUFED test set | MAP@5%: 0.296 -> 0.302 | Helped |
| include event-type information in the image importance prediction (ablation) | PR-CNN · CUFED test set | MAP@5%: 0.261 -> 0.302 | Helped |
| use piecewise ranking loss rather than conventional ranking SVM hinge loss (ablation) | CNN image-importance ranker · CUFED test set | MAP@5%: 0.266 -> 0.296 | Helped |
| add face heatmap-based scores (ablation) | Ensemble-CNN · Beach Trip | MAP@5%: +0.051 | Helped |
| add face heatmap-based scores (ablation) | Ensemble-CNN · Nature Trip | MAP@5%: +0.008 | Helped |
| add face heatmap-based scores (ablation) | Ensemble-CNN · Group Activity | MAP@5%: +0.003 | Helped |
| add face heatmap-based scores (ablation) | Ensemble-CNN · Halloween | MAP@5%: +0.000 | No clear effect |
| add face heatmap-based scores (ablation) | Ensemble-CNN · Museum | MAP@5%: -0.010 | Hurt |
| add face heatmap-based scores (ablation) | Ensemble-CNN · CUFED test set across event types | MAP: about 0.1% | No clear effect |
| use aesthetics scores instead of event-specific importance prediction | image ranking for event albums · CUFED test set | MAP@5%: 0.122 -> 0.139 | Hurt |
Closest papers by meaning
- Beyond Single Images: A Comprehensive Benchmark for Album-Level Vision-Language Understanding (CVPR 2026 · similarity 0.83)
- Fine-Grained Perception in Panoramic Scenes: A Novel Task, Dataset, and Method for Object Importance Ranking (AAAI 2025 · similarity 0.81)
- Event-Customized Image Generation (ICML 2025 · similarity 0.80)
- Doomed to Re-Annotate, Forever: The ImageNet Story (NeurIPS 2026 · similarity 0.80)
- SELECT: A Large-Scale Benchmark of Data Curation Strategies for Image Classification (NeurIPS 2024 · similarity 0.80)
- EventGPT: Event Stream Understanding with Multimodal Large Language Models (CVPR 2025 · similarity 0.80)
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.