Skip to main navigation Skip to search Skip to main content

Deconfounded Image Captioning: a causal retrospect

Research output: Contribution to journal › Article › Research › peer-review

Abstract

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.

Original languageEnglish
Pages (from-to)12996-13010
Number of pages14
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume45
Issue number11
DOIs
Publication statusPublished - 1 Nov 2023

Keywords

  • Causality
  • Correlation
  • Deconfounding
  • Image Captioning
  • Magnetic heads
  • Shape
  • Task analysis
  • the Backdoor Adjustment
  • the Front-Door Adjustment
  • Toy manufacturing industry
  • Training
  • Visualization

Cite this