去混淆图像描述：因果回溯
Deconfounded Image Captioning: A Causal Retrospect
AI总结：
本文提出基于因果推断的去混淆图像描述框架 DICv1.0，通过后门和前门调整缓解数据集偏差，并在 MS COCO 上取得显著性能提升。
摘要：
视觉-语言任务中的数据集偏差正成为阻碍我们社区进展的主要问题之一。现有解决方案缺乏关于现代图像描述器为何容易陷入数据集偏差的原则性分析。在本文中，我们提出一种新颖的视角：去混淆图像描述（Deconfounded Image Captioning, DIC），以找出该问题的答案，然后回溯现代神经图像描述器，最后提出一个 DIC 框架：DICv1.0，以减轻数据集偏差带来的负面影响。DIC 基于因果推断，其两个原则：后门调整和前门调整，帮助我们回顾先前研究并设计新的有效模型。特别地，我们展示了 DICv1.0 能够增强两种主流描述模型，并在具有挑战性的 MS COCO 数据集的 Karpathy 划分和在线划分上分别实现单模型 131.1 CIDEr-D 和 128.4 c40 CIDEr-D。有趣的是，DICv1.0 是我们因果回溯的自然推导，为图像描述开辟了有前景的方向。
英文摘要：
Dataset bias in vision-language tasks is becoming one of the main problems which hinders the progress of our community. Existing solutions lack a principled analysis about why modern image captioners easily collapse into dataset bias. In this paper, we present a novel perspective: Deconfounded Image Captioning (DIC), to find out the answer of this question, then retrospect modern neural image captioners, and finally propose a DIC framework: DICv1.0 to alleviate the negative effects brought by dataset bias. DIC is based on causal inference, whose two principles: the backdoor and front-door adjustments, help us review previous studies and design new effective models. In particular, we showcase that DICv1.0 can strengthen two prevailing captioning models and can achieve a single-model 131.1 CIDEr-D and 128.4 c40 CIDEr-D on Karpathy split and online split of the challenging MS COCO dataset, respectively. Interestingly, DICv1.0 is a natural derivation from our causal retrospect, which opens promising directions for image captioning.
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。
