Does Video Summarization Require Videos? Quantifying the Effectiveness of Language in Video Summarization

TL;DR

This paper introduces a language-only approach to video summarization that uses textual captions and transformer models, achieving competitive accuracy with high data efficiency and improved explainability.

Contribution

The authors propose a zero-shot, language-based video summarizer that eliminates the need for image data, demonstrating effective summarization solely from textual captions.

Findings

Language-only summarization reduces data processing significantly.

The approach achieves comparable accuracy to traditional methods.

Text modality enhances explainability and interpretability.

Abstract

Video summarization remains a huge challenge in computer vision due to the size of the input videos to be summarized. We propose an efficient, language-only video summarizer that achieves competitive accuracy with high data efficiency. Using only textual captions obtained via a zero-shot approach, we train a language transformer model and forego image representations. This method allows us to perform filtration amongst the representative text vectors and condense the sequence. With our approach, we gain explainability with natural language that comes easily for human interpretation and textual summaries of the videos. An ablation study that focuses on modality and data compression shows that leveraging text modality only effectively reduces input data processing while retaining comparable results.