期刊
JOURNAL OF REAL-TIME IMAGE PROCESSING
卷 18, 期 6, 页码 2319-2329出版社
SPRINGER HEIDELBERG
DOI: 10.1007/s11554-021-01124-9
关键词
Fire detection; YOLOv4; MobileNetV3; Model compression; Channel-level sparsity
类别
资金
- National Natural Science Foundation of China [61861001]
- Postgraduate Innovation Project of North Minzu University [YCX20111]
This study proposes a deep learning fire recognition algorithm based on model compression and lightweight requirements, utilizing the MobileNetV3 model to simplify the conventional network structure and improve detection accuracy through knowledge distillation. Experimental results demonstrate significant advantages in reducing model parameters and inference time compared to existing algorithms.
To meet the needs of embedded intelligent forest fire monitoring systems using an unmanned aerial vehicles (UAV), a deep learning fire recognition algorithm based on model compression and lightweight requirements is proposed in this study. The algorithm for the lightweight MobileNetV3 model was developed to reduce the complexity of the conventional YOLOv4 network structure. The redundant channels are eliminated through channel-level sparsity-induced regularization. The knowledge distillation algorithm is used to improve the detection accuracy of the pruned model. The experimental results reveal that the number of model parameters for the proposed architecture is only 2.64 million-compared with YOLOv4, this represents a reduction of nearly 95.87%. The inference time decreased from 153.8 to 37.4 ms, a reduction of nearly 75.68%. Our approach shows the advantages of a model with a smaller number of parameters, low memory requirements and fast inference speed compared with existing algorithms. The method presented in this paper is specifically tailored for use as a deep learning forest fire monitoring system on a UAV platform.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据