Scholarlink

<p>Eye state detection plays a crucial role in human–computer interaction, fatigue detection, psychological state analysis, and other fields. In recent years, deep learning-based methods have achieved good results in eye state detection. However, eye state detection algorithms still suffer from the problem that the detection effect is easily affected by occlusion and head deflection. To address these problems, this paper improves on single-shot multi-box detector. An eye state detection model based on feature fusion and attention mechanism SSD (FFA-SSD) is proposed. Firstly, a feature fusion module is designed to enhance the semantic information of shallow features. Then, an improved attention mechanism module is introduced to utilize spatial attention and channel attention to suppress background noise and overcome the effect of occlusion on detection. In addition, a feature enhancement module is proposed to effectively enhance the feature representation of the network and improve the detection accuracy in the case of head deflection. Finally, the proposed method is tested on the public dataset Closed Eyes In The Wild (CEW). The experimental results show that the mAP of the method in this paper can reach 97.64%, which significantly improves the detection accuracy for occlusion and complex environments. Our code and models will be available at <a href="https://github.com/daiweikun/FFA-SSD">https://github.com/daiweikun/FFA-SSD</a>.</p>

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

Eye state detection based on feature fusion and attention single-shot multi-box detector

  • 期刊论文

  • 发表日期：

    2025年8月24日

查看全文
  • Weikun Dai,
  • Jianbin Xiong,
  • Xiangjun Dong,
  • Qi Wang,
  • Jianxiang Yang,
  • Wulue Zhang,
  • Qianguang Zhang,
  • Jing Wang,
  • Jinchang Ren

摘要

Eye state detection plays a crucial role in human–computer interaction, fatigue detection, psychological state analysis, and other fields. In recent years, deep learning-based methods have achieved good results in eye state detection. However, eye state detection algorithms still suffer from the problem that the detection effect is easily affected by occlusion and head deflection. To address these problems, this paper improves on single-shot multi-box detector. An eye state detection model based on feature fusion and attention mechanism SSD (FFA-SSD) is proposed. Firstly, a feature fusion module is designed to enhance the semantic information of shallow features. Then, an improved attention mechanism module is introduced to utilize spatial attention and channel attention to suppress background noise and overcome the effect of occlusion on detection. In addition, a feature enhancement module is proposed to effectively enhance the feature representation of the network and improve the detection accuracy in the case of head deflection. Finally, the proposed method is tested on the public dataset Closed Eyes In The Wild (CEW). The experimental results show that the mAP of the method in this paper can reach 97.64%, which significantly improves the detection accuracy for occlusion and complex environments. Our code and models will be available at https://github.com/daiweikun/FFA-SSD.