arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.01860cs.CVcs.CL
收藏

L2C:描述视觉差异需要个体的语义理解

L2C: Describing Visual Differences Needs Semantic Understanding of Individuals

UC San Diego(加州大学圣地亚哥分校);UC Santa Cruz(加州大学圣克鲁兹分校);UC Santa Barbara(加州大学圣巴巴拉分校)

An Yan, Xin Eric Wang, Tsu-Jui Fu, William Yang Wang

摘要:

语言和视觉的最新进展推动了对单个图像的字幕描述图像对之间的视觉差异的研究。假设有两幅图像I_1和I_2,任务是生成一个描述W_{1,2}来比较它们,现有的方法直接对{ I_1,I_2 } -> W_{1,2}映射建模,而不需要个体的语义理解。在本文中,我们介绍了一个学习比较(L2 C)模型,它学习理解这两个图像的语义结构,并比较它们,同时学习描述每一个。我们证明,L2 C受益于明确的语义表示和单图像标题之间的比较,并更好地推广新的测试图像对。它在Birds-to-Words数据集的自动评估和人工评估方面都优于基线。

英文摘要:

Recent advances in language and vision push forward the research of captioning a single image to describing visual differences between image pairs. Suppose there are two images, I_1 and I_2, and the task is to generate a description W_{1,2} comparing them, existing methods directly model { I_1, I_2 } -> W_{1,2} mapping without the semantic understanding of individuals. In this paper, we introduce a Learning-to-Compare (L2C) model, which learns to understand the semantic structures of these two images and compare them while learning to describe each one. We demonstrate that L2C benefits from a comparison between explicit semantic representations and single-image captions, and generalizes better on the new testing image pairs. It outperforms the baseline on both automatic evaluation and human evaluation for the Birds-to-Words dataset.

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。

↑