Cross-Modal Similarity-Based Curriculum Learning for Image Captioning

TL;DR

This paper introduces a cross-modal similarity-based curriculum learning method for image captioning, which improves performance and generalization by adaptively measuring difficulty using a pretrained vision-language model.

Contribution

It proposes a novel difficulty measurement for image captioning based on cross-modal similarity, enhancing training efficiency and model generalization without extra costs.

Findings

Achieves superior performance on COCO and Flickr30k datasets.

Demonstrates improved generalization to unseen data.

Maintains competitive convergence speed.

Abstract

Image captioning models require the high-level generalization ability to describe the contents of various images in words. Most existing approaches treat the image-caption pairs equally in their training without considering the differences in their learning difficulties. Several image captioning approaches introduce curriculum learning methods that present training data with increasing levels of difficulty. However, their difficulty measurements are either based on domain-specific features or prior model training. In this paper, we propose a simple yet efficient difficulty measurement for image captioning using cross-modal similarity calculated by a pretrained vision-language model. Experiments on the COCO and Flickr30k datasets show that our proposed approach achieves superior performance and competitive convergence speed to baselines without requiring heuristics or incurring additional training costs. Moreover, the higher model performance on difficult examples and unseen data also demonstrates the generalization ability.