MeaCap: Memory-Augmented Zero-shot Image Captioning

TL;DR

MeaCap is a novel memory-augmented framework for zero-shot image captioning that enhances caption relevance and reduces hallucinations by retrieving and filtering key concepts related to the image before generating captions.

Contribution

The paper introduces MeaCap, which incorporates a textual memory and a retrieve-then-filter module to improve zero-shot image captioning performance and caption quality.

Findings

Achieves state-of-the-art results on zero-shot IC benchmarks.

Generates more accurate, concept-centered captions with fewer hallucinations.

Effectively integrates memory-augmented retrieval with language modeling.

Abstract

Zero-shot image captioning (IC) without well-paired image-text data can be divided into two categories, training-free and text-only-training. Generally, these two types of methods realize zero-shot IC by integrating pretrained vision-language models like CLIP for image-text similarity evaluation and a pre-trained language model (LM) for caption generation. The main difference between them is whether using a textual corpus to train the LM. Though achieving attractive performance w.r.t. some metrics, existing methods often exhibit some common drawbacks. Training-free methods tend to produce hallucinations, while text-only-training often lose generalization capability. To move forward, in this paper, we propose a novel Memory-Augmented zero-shot image Captioning framework (MeaCap). Specifically, equipped with a textual memory, we introduce a retrieve-then-filter module to get key concepts that are highly related to the image. By deploying our proposed memory-augmented visual-related fusion score in a keywords-to-sentence LM, MeaCap can generate concept-centered captions that keep high consistency with the image with fewer hallucinations and more world-knowledge. The framework of MeaCap achieves the state-of-the-art performance on a series of zero-shot IC settings. Our code is available at https://github.com/joeyz0z/MeaCap.