Scholarlink

<p>With the growing popularity of smartphones, the demand for mobile applications has increased. However, developing mobile apps can be challenging. Developers often use online videos to help them. Finding an appropriate screencast can delay the development process. Although researchers have tried to extract valuable information from screencasts using various techniques, there is still no accurate method for summarization. Therefore, this research proposes a method to summarize mobile screencasts by considering all relevant information. It enables programmers to quickly assess the relevance of a screencast. This method extracts frames and removes redundant or noninformative ones. It then uses MobileNetV2 and the Histogram of Oriented Gradients (HOG) method to extract features from the remaining frames. Finally, K-means clustering groups the frames and selects key ones for the summary. The proposed method outperforms VSUMM, GVSUM, VISFF, and SVS_MCO in F1-score by 7.37%, 16.9%, 15.25%, and 7.8%, respectively. It also provides summaries that help programmers save time and effort in identifying relevant screencasts.</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

Summarizing Mobile Programming Screencasts

  • 期刊论文

  • 发表日期：

    2025年7月10日

查看全文
  • Fatemeh Khoshnoud,
  • Mohammad R. Moosavi,
  • Neda Azouji

摘要

With the growing popularity of smartphones, the demand for mobile applications has increased. However, developing mobile apps can be challenging. Developers often use online videos to help them. Finding an appropriate screencast can delay the development process. Although researchers have tried to extract valuable information from screencasts using various techniques, there is still no accurate method for summarization. Therefore, this research proposes a method to summarize mobile screencasts by considering all relevant information. It enables programmers to quickly assess the relevance of a screencast. This method extracts frames and removes redundant or noninformative ones. It then uses MobileNetV2 and the Histogram of Oriented Gradients (HOG) method to extract features from the remaining frames. Finally, K-means clustering groups the frames and selects key ones for the summary. The proposed method outperforms VSUMM, GVSUM, VISFF, and SVS_MCO in F1-score by 7.37%, 16.9%, 15.25%, and 7.8%, respectively. It also provides summaries that help programmers save time and effort in identifying relevant screencasts.