Scholarlink

Detection of anomalies has a major role to play in the field of video analytics. It focuses on identifying abnormal events in captured surveillance or Dashcam footage, by leveraging the advantages of data-driven algorithms. The proposed algorithm enhances and optimizes the existing problem of traffic anomaly detection by combining the effective feature extraction of DenseNet-201 model, the spatiotemporal processing prowess of the Video Swin Transformer, and the temporal modeling skills of Long Short-Term Memory (LSTM) networks. The method discussed in this paper integrates genetic algorithm with the hybrid model to enhance the results. Exhaustive experimentation on a meticulously curated DoTA dataset reveals that the novel technique is able to attain significant performance gains over current methods.

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

Traffic Anomaly Detection Using Novel GEN-Sembling Technique

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Harkiran Kaur,
  • Vatsla Zharaik

摘要

Detection of anomalies has a major role to play in the field of video analytics. It focuses on identifying abnormal events in captured surveillance or Dashcam footage, by leveraging the advantages of data-driven algorithms. The proposed algorithm enhances and optimizes the existing problem of traffic anomaly detection by combining the effective feature extraction of DenseNet-201 model, the spatiotemporal processing prowess of the Video Swin Transformer, and the temporal modeling skills of Long Short-Term Memory (LSTM) networks. The method discussed in this paper integrates genetic algorithm with the hybrid model to enhance the results. Exhaustive experimentation on a meticulously curated DoTA dataset reveals that the novel technique is able to attain significant performance gains over current methods.