Scholarlink

In this study, YAPN (Yaw Angle Prediction Net), a deep learning framework for pose estimation is presented. This framework uses RGB images captured in real-world traffic environments to estimate the yaw angle, representing the pose of vehicles within the scene. PEN (Part Encoding Network) is used to detect the individual parts of the vehicle and the yaw angle predictor estimates the yaw angle of the vehicle. In real-world scenarios, YAPN shows great effectiveness with an average prediction error of less than 3° and gives an accuracy of 96% for predictions within 10°. The framework’s capacity to generalize across diverse environments, manage data restrictions, and satisfy hardware requirements remain significant obstacles.

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

Single Image Based Vehicle Pose Estimation in Complex Traffic Environments

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shashidhar Angadi,
  • Suresh G. Kini,
  • Rajesh Katagar,
  • Kaushik Mallibhat,
  • Prabha Nissimagoudar

摘要

In this study, YAPN (Yaw Angle Prediction Net), a deep learning framework for pose estimation is presented. This framework uses RGB images captured in real-world traffic environments to estimate the yaw angle, representing the pose of vehicles within the scene. PEN (Part Encoding Network) is used to detect the individual parts of the vehicle and the yaw angle predictor estimates the yaw angle of the vehicle. In real-world scenarios, YAPN shows great effectiveness with an average prediction error of less than 3° and gives an accuracy of 96% for predictions within 10°. The framework’s capacity to generalize across diverse environments, manage data restrictions, and satisfy hardware requirements remain significant obstacles.