arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.03567cs.CV
收藏

通过循环一致的文本到图像生成评估图像描述

Evaluating Image Caption via Cycle-consistent Text-to-Image Generation

ShanghaiTech University(上海科技大学);Alibaba Group(阿里巴巴集团)

Tianyu Cui, Jinbin Bai, Guo-Hua Wang, Qing-Guo Chen, Zhao Xu, Weihua Luo, Kaifu Zhang, Ye Shi

AI总结:

针对图像描述评估中参考描述成本高及跨模态指标受模态鸿沟影响的问题,本文提出无参考的 CAMScore,通过文本到图像生成与三级图像比较提升其与人类判断的相关性。

摘要:

评估图像描述通常依赖参考描述,而参考描述获取成本高昂,并表现出显著的多样性和主观性。尽管已有研究提出无参考评估指标,但大多数指标聚焦于描述与图像之间的跨模态评估。近期研究表明,基于对比学习的多模态系统在表征中普遍存在模态鸿沟,这会削弱 CLIPScore 等跨模态指标的可靠性。本文提出 CAMScore,这是一种用于图像描述模型的循环式无参考自动评估指标。为规避上述模态鸿沟,CAMScore 利用文本到图像模型根据描述生成图像,随后将这些生成图像与原始图像进行评估。此外,为提供细粒度信息以实现更全面的评估,我们为 CAMScore 设计了一个三级评估框架,涵盖像素级、语义级和目标级视角。在多个基准数据集上的大量实验结果表明,与现有的基于参考和无参考指标相比,CAMScore 与人类判断具有更强的相关性,证明了该框架的有效性。

英文摘要:

Evaluating image captions typically relies on reference captions, which are costly to obtain and exhibit significant diversity and subjectivity. While reference-free evaluation metrics have been proposed, most focus on cross-modal evaluation between captions and images. Recent research has revealed that the modality gap generally exists in the representation of contrastive learning-based multi-modal systems, undermining the reliability of cross-modality metrics like CLIPScore. In this paper, we propose CAMScore, a cyclic reference-free automatic evaluation metric for image captioning models. To circumvent the aforementioned modality gap, CAMScore utilizes a text-to-image model to generate images from captions and subsequently evaluates these generated images against the original images. Furthermore, to provide fine-grained information for a more comprehensive evaluation, we design a three-level evaluation framework for CAMScore that encompasses pixel-level, semantic-level, and objective-level perspectives. Extensive experiment results across multiple benchmark datasets show that CAMScore achieves a superior correlation with human judgments compared to existing reference-based and reference-free metrics, demonstrating the effectiveness of the framework.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑