| 基于深度学习的图书破损检测方法研究 |
 点此下载全文 |
| 引用本文：张志颖1，刘竟1,2，许梦怡3，郭剑明1.基于深度学习的图书破损检测方法研究[J].计算技术与自动化,2025,(3):141-146 |
| 摘要点击次数: 298 |
| 全文下载次数: 209 |
|
|
| 中文摘要:破损图书管理和维护是图书馆中的一项重要工作。在进行破损图书筛查时，传统的人工筛查效率低下且易受主观因素影响，难以满足现代图书馆的需求。因此，提出了一种基于深度学习的图书破损检测模型。本文以YOLOv5为基础模型，引入CBAM注意力机制、Meta-ACON激活函数和加权双向特征金字塔网络（BiFPN）对其进行改进。所提模型的mAP@0.5达到了94.89%，其中破损图书类别的AP达到了94.59%，与YOLOv5s相比提高了2.42%。该模型可以在不同背景下提供良好的检测性能，有助于提高筛查破损图书的效率，及时维护破损的图书。 |
| 中文关键词:深度学习 目标检测 YOLOv5 破损图书检测 CBAM Meta-ACON BiFPN |
| |
| Research on Book Damage Detection Method Based on Deep Learning |
|
|
| Abstract:The management and maintenance of damaged books has always been an important work in the library. In the detecting of damaged books, the traditional manual detecting is inefficient and susceptible to subjective factors, which makes it difficult to meet the needs of modern libraries. Therefore, we proposed a book damage detection model based on deep learning. We used YOLOv5 as a base model and improved it by introducing convolutional block attention module (CBAM), Meta-ACON activation function and bi-directional feature pyramid network (BiFPN). The mAP@0.5 of the proposed model reaches 94.89%. Specially, the AP of the damaged book reaches 94.59%, which is 2.42% higher than that of YOLOv5s. This model can provide good detection performance in different backgrounds, which helps to improve the efficiency of screening damaged books and timely maintenance of damaged books. |
| keywords:deep learning object detection YOLOv5 book damage detection CBAM Meta-ACON BiFPN |
| 查看全文 查看/发表评论 下载pdf阅读器 |
|
|
|
位访问者,今日访问
次