Scholarlink

This research presents a novel approach to video summarization (VS) for efficiently extracting keyframes, reducing video redundancy while preserving essential content. The method combines deep learning and clustering techniques to generate concise video summaries by analyzing videos frame by frame. Using a dual convolutional neural network (CNN), we extract deep-level features. K-means clustering is then applied to group the feature descriptors of the video frames into keyframes and non-keyframes. This K-means clustering-based VS (KVS) method effectively selects the most relevant frames from the extracted features. Our proposed KVS approach outperforms existing VS techniques, achieving an average F-score of 73.1 and 76.3 on two benchmark datasets, Open Video and YouTube, respectively, demonstrating its superior ability to produce informative video summaries.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Static Video Summarization Using Transfer Learning and Clustering

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shamal Kashid,
  • Lalit K. Awasthi,
  • Krishan Berwal,
  • Parul Saini

摘要

This research presents a novel approach to video summarization (VS) for efficiently extracting keyframes, reducing video redundancy while preserving essential content. The method combines deep learning and clustering techniques to generate concise video summaries by analyzing videos frame by frame. Using a dual convolutional neural network (CNN), we extract deep-level features. K-means clustering is then applied to group the feature descriptors of the video frames into keyframes and non-keyframes. This K-means clustering-based VS (KVS) method effectively selects the most relevant frames from the extracted features. Our proposed KVS approach outperforms existing VS techniques, achieving an average F-score of 73.1 and 76.3 on two benchmark datasets, Open Video and YouTube, respectively, demonstrating its superior ability to produce informative video summaries.