Skip to content
    Flow-alerta.cab

    CIDEr

    AI fundamentals & modelsConsensus-based Image Description EvaluationCIDEr metric
    Editorially reviewed on: 1 source
    Your reading view

    The short answer · 16 words

    In short

    Automatic evaluation metric for image descriptions that compares a generated caption with several human reference descriptions.

    Flow-alerta.cab (2026): CIDEr. https://flow-alerta.cab/en/glossar/cider. Editorially reviewed on: 09 October 2026.

    How much detail would you like?

    Practical

    Relevant for the evaluation of image captioning systems and multimodal models, if reference descriptions are available.

    Example scenario: Two image description models are compared on the same data set with several human reference captions per image.

    Use only for tasks with suitable reference descriptions and supplemented by human and task-specific quality checks.

    At a glance

    Core
    CIDEr evaluates how strongly n-grams of a generated image description match the consensus of multiple human reference descriptions. For this purpose, n-grams are weighted with TF-IDF, so that characteristic formulations count more than very common words. The metric evaluates text similarity to references, not automatically image truth or general language quality.
    Use
    Relevant for the evaluation of image captioning systems and multimodal models, if reference descriptions are available.
    Example
    Example scenario: Two image description models are compared on the same data set with several human reference captions per image.
    Limits
    Understand CIDEr as a general image quality metric; compare values from different data sets directly; equate high similarity with factual fidelity.

    In context

    BLEU Score

    Automatic metric for evaluating machine-generated translations based on the match of n-grams with reference translations.

    BERTScore

    Automatic evaluation metric for generated text that compares token similarities based on contextual embeddings instead of just exact word matches.

    Questions about this term

    Does CIDEr measure image quality?

    No.

    Do you need reference descriptions?

    Yes.

    Is a higher value fundamentally better?

    Within the same evaluation typically yes.

    All details and background

    CIDEr evaluates how strongly n-grams of a generated image description match the consensus of multiple human reference descriptions. For this purpose, n-grams are weighted with TF-IDF, so that characteristic formulations count more than very common words. The metric evaluates text similarity to references, not automatically image truth or general language quality.

    ELI5 – Explained simply

    Several people describe the same picture. CIDEr checks how well the automatically generated description fits this common description pattern.

    Why it matters in practice

    Relevant for the evaluation of image captioning systems and multimodal models, if reference descriptions are available.

    Example

    Example scenario: Two image description models are compared on the same data set with several human reference captions per image.

    Pitfalls and limits

    • Understand CIDEr as a general image quality metric; compare values from different data sets directly; equate high similarity with factual fidelity.

    Origin & history

    CIDEr was proposed in 2014/2015 as a consensus-based evaluation metric for automatic image descriptions.

    Use in mid-sized companies

    • Use only for tasks with suitable reference descriptions and supplemented by human and task-specific quality checks.

    Related terms

    Matching Flow-alerta service

    AI opportunity analysis & roadmapAI consulting: AI potential analysis & roadmap on tasks, data and prerequisites of a possible first AI use case.

    Matching tool / use case

    View use casesUse Case: Comparison of two image captioning models on a fixed test dataset.

    Sources

    • arxiv.orgVedantam, Zitnick & Parikh, CIDEr: Consensus-based Image Description Evaluation

    Quellenprüfung: 09.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.

    Editorially reviewed on:

    Next step

    Put knowledge into practice

    Discuss your projectView use cases