Video Summarization using Denoising Diffusion Probabilistic Model

TL;DR

This paper introduces a novel generative video summarization method using Denoising Diffusion Probabilistic Models, which reduces annotation noise impact and improves generalization over traditional discriminative approaches.

Contribution

The paper proposes a diffusion-based generative framework for video summarization that is more robust to subjective annotation noise and overfitting, with an unsupervised pretraining strategy.

Findings

Outperforms existing methods on TVSum, SumMe, and FPVSum datasets.

Demonstrates robustness to annotation noise and limited data.

Achieves strong generalization in video summarization tasks.

Abstract

Video summarization aims to eliminate visual redundancy while retaining key parts of video to construct concise and comprehensive synopses. Most existing methods use discriminative models to predict the importance scores of video frames. However, these methods are susceptible to annotation inconsistency caused by the inherent subjectivity of different annotators when annotating the same video. In this paper, we introduce a generative framework for video summarization that learns how to generate summaries from a probability distribution perspective, effectively reducing the interference of subjective annotation noise. Specifically, we propose a novel diffusion summarization method based on the Denoising Diffusion Probabilistic Model (DDPM), which learns the probability distribution of training data through noise prediction, and generates summaries by iterative denoising. Our method is more resistant to subjective annotation noise, and is less prone to overfitting the training data than discriminative methods, with strong generalization ability. Moreover, to facilitate training DDPM with limited data, we employ an unsupervised video summarization model to implement the earlier denoising process. Extensive experiments on various datasets (TVSum, SumMe, and FPVSum) demonstrate the effectiveness of our method.