AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Describing Like Humans: On Diversity in Image Captioning

Qingzhong Wang, Antoni B. Chan

CVPR 2019 · Vision-language and multimodal › Human and embodied modeling · approach: Neural image captioning · 96 citations (OpenAlex)

Check 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 changedWhereMeasuredEffect
Use Self-CIDEr rather than mBLEU-mix to measure caption-set diversityCaption diversity evaluation · Human diversity judgmentPearson correlation: 0.585 -> 0.616Improved
Use Self-CIDEr rather than mBLEU-mix to measure caption-set diversityCaption diversity evaluation · Human diversity judgmentSpearman correlation: 0.575 -> 0.617Improved
Use LSA rather than Self-CIDEr to measure caption-set diversityCaption diversity evaluation · Human diversity judgmentAverage per-image Spearman correlation: 0.674 -> 0.678Improved
Balance cross-entropy loss with CIDEr reward during RL trainingAtt2in and FC captioners · MSCOCO Karpathy test splitDiversity and accuracy tradeoffImproved
Increase the CIDEr reward weight (ablation)Att2in and FC captioners · MSCOCO Karpathy test splitDiversityHurt
Increase the CIDEr reward weight (ablation)Att2in and FC captioners · MSCOCO Karpathy test splitAccuracyImproved
Add retrieval reward to the combined loss (ablation)Att2in and FC captioners · MSCOCO Karpathy test splitDiversity and accuracyImproved
Use a very large retrieval-reward weight (ablation)Att2in and FC captioners · MSCOCO Karpathy test splitCaption qualityHurt
Use random sampling instead of a deterministic single caption (ablation)Captioning models · MSCOCO Karpathy test splitDiversityImproved
Use random sampling instead of a deterministic single caption (ablation)Captioning models · MSCOCO Karpathy test splitAccuracyHurt
Use randomly cropped images to generate captions (ablation)Captioning models · MSCOCO Karpathy test splitAccuracyImproved
Use retrieval reward rather than CIDEr reward (ablation)Att2in and FC captioners · MSCOCO Karpathy test splitDiversityImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy