Recognizing characters and predicting speakers of dialogue are critical for comic processing tasks, such as voice generation or translation. However, because characters vary by comic title, supervised learning approaches like training character classifiers which require specific annotations for each comic title are infeasible. This motivates us to propose a novel zero-shot approach, allowing machines to identify characters and predict speaker names based solely on unannotated comic images. In spite of their importance in real-world applications, these task have largely remained unexplored due to challenges in story comprehension and multimodal integration. Recent large language models (LLMs) have shown great capability for text understanding and reasoning, while their application to multimodal content analysis is still an open problem. To address this problem, we propose an iterative multimodal framework, the first to employ multimodal information for both character identification and speaker prediction tasks. Our experiments demonstrate the effectiveness of the proposed framework, establishing a robust baseline for these tasks. Furthermore, since our method requires no training data or annotations, it can be used as-is on any comic series.
We pioneer in revealing the potential of large language models (LLMs) for comics analysis and propose a novel method that integrates text and image modalities. To address the challenges of integrating LLMs with other modules and enhance the machine's comprehension of comics, we introduce an iterative framework. We merge text-based LLM predictions with image-based classifiers, and alternately refine each module using results from the other.
In the main evaluation, we omit the steps of object detection, OCR, and character name extraction in data preprocessing. That is, we simplify the tasks of speaker prediction and character identification to classify the character labels for the character and text regions. The experimental results suggest that our iterative process is effective, particularly when the prediction of relationships between text and character regions is accurate.
To validate the effectiveness of our proposed iterative multimodal fusion method, we conducted evaluations in two aspects: Unimodal vs. Multimodal and One-Step vs. Iterative.
@inproceedings{li2024zeroshot,
title={Zero-shot character identification and speaker prediction in comics via iterative multimodal fusion},
author={Li, Yingxuan and Hinami, Ryota and Aizawa, Kiyoharu and Matsui, Yusuke},
booktitle={Proceedings of the ACM International Conference on Multimedia},
year={2024}
}