Scholarlink

The rapid expansion of ride-hailing services has heightened safety concerns, with incidents of harassment and violence requiring proactive solutions. Traditional measures like emergency buttons lack real-time intervention capabilities. This paper explores two distinct AI-powered approaches for in-vehicle violence detection using their MoLa InCar AR dataset (1) YOLOv8 for real-time object detection, achieving 99.46% mAP@50 in localizing violent actions, and (2) TSM-ResNet50 for temporal action recognition, attaining 93.62% accuracy in classifying aggressive behaviors. Both systems trigger automated alerts and securely store evidence upon detection. Evaluations demonstrate their complementary strengths—YOLOv8 excels in spatial precision, while TSM-ResNet50 captures temporal dynamics—offering versatile solutions for ride-hailing safety. The paper underscores the potential of computer vision to enhance transportation security through parallel, specialized models.”

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

AI Models for Real-Time Violence Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mohannad Waleed,
  • Mariam Yasser,
  • Mayada Magdy,
  • Maryam Elkady,
  • Mariam Ayman,
  • Tasneem Hesham,
  • Abdelrahman Adel,
  • Youssef Sherif,
  • Mostafa Alazzaly,
  • Ahmed Abdelatif,
  • Khaled Abdelsalam

摘要

The rapid expansion of ride-hailing services has heightened safety concerns, with incidents of harassment and violence requiring proactive solutions. Traditional measures like emergency buttons lack real-time intervention capabilities. This paper explores two distinct AI-powered approaches for in-vehicle violence detection using their MoLa InCar AR dataset (1) YOLOv8 for real-time object detection, achieving 99.46% mAP@50 in localizing violent actions, and (2) TSM-ResNet50 for temporal action recognition, attaining 93.62% accuracy in classifying aggressive behaviors. Both systems trigger automated alerts and securely store evidence upon detection. Evaluations demonstrate their complementary strengths—YOLOv8 excels in spatial precision, while TSM-ResNet50 captures temporal dynamics—offering versatile solutions for ride-hailing safety. The paper underscores the potential of computer vision to enhance transportation security through parallel, specialized models.”