Scholarlink

<p>Video captioning aims to generate coherent and semantically accurate descriptions for input videos. Despite recent advances, challenges remain in modeling temporal scene evolution and capturing fine-grained contextual details. We propose the Contextual Anchor Semantic Enhanced (CASE) framework, incorporating temporal scene understanding and dynamic feature aggregation to enhance global semantic modeling. Our Contextual Anchor-Guided Encoder captures temporal dependencies, while the Frame-level Global Aggregator dynamically integrates motion and appearance cues. Extensive experiments demonstrate that our method achieves state-of-the-art performance, delivering consistent improvements across most evaluation metrics on both MSVD and MSR-VTT benchmarks.</p>

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

Enhancing video captioning with contextual anchor-guided semantic modeling

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Zhonghua Li,
  • Lina Yang,
  • Xichun Li,
  • Thomas Wu,
  • Yifeng Tan,
  • Yijun Chen

摘要

Video captioning aims to generate coherent and semantically accurate descriptions for input videos. Despite recent advances, challenges remain in modeling temporal scene evolution and capturing fine-grained contextual details. We propose the Contextual Anchor Semantic Enhanced (CASE) framework, incorporating temporal scene understanding and dynamic feature aggregation to enhance global semantic modeling. Our Contextual Anchor-Guided Encoder captures temporal dependencies, while the Frame-level Global Aggregator dynamically integrates motion and appearance cues. Extensive experiments demonstrate that our method achieves state-of-the-art performance, delivering consistent improvements across most evaluation metrics on both MSVD and MSR-VTT benchmarks.