Artificial Intelligence
CIDEr
Also known as:
CIDEr Score
Consensus-based Image Description Evaluation
Updated: 2/12/2026
A metric for image captioning that measures TF-IDF-weighted n-gram similarity.
Quick Summary
CIDEr uses TF-IDF-weighted n-gram comparison for image captioning – standard on COCO.
Explanation
CIDEr down-weights frequent words (TF-IDF) and rewards description-specific terms.
Marketing Relevance
CIDEr is the standard metric for image captioning benchmarks (COCO Captions).
Common Pitfalls
Favors "average" descriptions. Creative captions receive lower scores.
Origin & History
Vedantam et al. (2015) developed CIDEr for the COCO captioning challenge.
Comparisons & Differences
CIDEr vs. BLEU Score
BLEU uses equally weighted n-grams; CIDEr uses TF-IDF weighting.