Scholarlink

In this research, we analyse helmet detection with transformer-based and convolutional methods. First, we used a CNN-based classifier for helmet classification and YOLOv8 for detection. We then used a refined Swin Transformer to expand our investigation, assessing both models on inference time, precision, recall and F1-score. According to our comparative analysis, transformer-based models can maintain competitive inference efficiency while outperforming CNNs in accuracy.

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

Fine-Tuning Swin Transformer for Helmet Detection: A Comparative Study with CNNs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Vineet Kumar,
  • Tushar Agrawal,
  • Mahesh Jangid

摘要

In this research, we analyse helmet detection with transformer-based and convolutional methods. First, we used a CNN-based classifier for helmet classification and YOLOv8 for detection. We then used a refined Swin Transformer to expand our investigation, assessing both models on inference time, precision, recall and F1-score. According to our comparative analysis, transformer-based models can maintain competitive inference efficiency while outperforming CNNs in accuracy.