Scholarlink

With the acceleration of urbanization and the increasing maturity of intelligent transportation systems, pedestrian detection has become one of the most important tasks in computer vision. In response to the problem of poor performance of traditional methods in complex backgrounds and occlusion situations, this paper proposes an innovative pedestrian detection algorithm - YOLO-GNN, which is an improved version of the YOLOv4 algorithm and combines advanced technology of graph neural networks. YOLO-GNN effectively solves the problem of pedestrian occlusion and improves its detection ability in complex street scenes by utilizing the contextual information within the frame and the node information aggregation ability of GNN. Our experimental results indicate that YOLO-GNN achieved 81.2% mAP on the Caltech dataset, a significant improvement of 2.4% compared to baseline YOLOv4.

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

YOLO-GNN: An Improved GNN Based YOLOv4 Algorithm for Street Pedestrian Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Furong Peng,
  • Yuyu Chen

摘要

With the acceleration of urbanization and the increasing maturity of intelligent transportation systems, pedestrian detection has become one of the most important tasks in computer vision. In response to the problem of poor performance of traditional methods in complex backgrounds and occlusion situations, this paper proposes an innovative pedestrian detection algorithm - YOLO-GNN, which is an improved version of the YOLOv4 algorithm and combines advanced technology of graph neural networks. YOLO-GNN effectively solves the problem of pedestrian occlusion and improves its detection ability in complex street scenes by utilizing the contextual information within the frame and the node information aggregation ability of GNN. Our experimental results indicate that YOLO-GNN achieved 81.2% mAP on the Caltech dataset, a significant improvement of 2.4% compared to baseline YOLOv4.