Text-only Synthesis for Image Captioning

TL;DR

This paper introduces ToCa, a text-only synthesis method for image captioning that reduces reliance on large annotated datasets by generating diverse captions with minimal human effort, improving zero-shot and data-efficient performance.

Contribution

We propose a novel text-only synthesis approach that decomposes captions into structures and lexical words, enabling large language models to generate diverse, high-quality captions with fewer resources.

Findings

Achieves nearly 5 CIDEr improvement in zero-shot cross-domain captioning.

Over 20 CIDEr increase in data-efficient captioning.

Demonstrates strong generalizability and transferability across scenarios.

Abstract

From paired image-text training to text-only training for image captioning, the pursuit of relaxing the requirements for high-cost and large-scale annotation of good quality data remains consistent. In this paper, we propose Text-only Synthesis for Image Captioning (ToCa), which further advances this relaxation with fewer human labor and less computing time. Specifically, we deconstruct caption text into structures and lexical words, which serve as the fundamental components of the caption. By combining different structures and lexical words as inputs to the large language model, massive captions that contain various patterns of lexical words are generated. This method not only approaches the target domain but also surpasses it by generating new captions, thereby enhancing the zero-shot generalization ability of the model. Considering the different levels of data access in the real world, we define three synthesis scenarios: cross-domain synthesis, in-domain synthesis, and data-efficient synthesis. Experiments in these scenarios demonstrate the generalizability, transferability and practicability of ToCa with a nearly 5 CIDEr improvement for zero-shot cross-domain captioning and a maximum increase of over 20 CIDEr for data-efficient captioning.