基于MobileNet-CBAM模型的轻量级鸟类识别方法
Lightweight bird recognition method based on mobileNet-CBAM model
成迪 1雷毅 1张成军 1陈亚当 1刘文杰 1余文斌1
作者信息
- 1. 南京信息工程大学,江苏大气环境和装备技术协同创新中心,南京 210044;南京信息工程大学计算机学院,南京 210044;南京信息工程大学数字取证教育部工程研究中心,南京 210044
- 折叠
摘要
对于鸟类识别任务,当前一些深度学习网络模型的效果并不理想,针对现有模型的痛点,提出了一种在识别精度、计算性能、资源消耗和实用性上更具优势的轻量级鸟类识别方法.该方法基于轻量级 MobileNet 模型,融合了 卷积注意力机制(Convolutional Block Attention Module,CBAM),在减少模型复杂度的同时增强了局部特征的提取能力,与主流模型的对比实验验证了该模型的有效性.此外,该模型能够应用于机器人、便携式手机等硬件平台,扩展了应用场景,并添加数据库来进行辅助计算,减少额外开销,兼具便携性和实用性.
Abstract
For bird recognition tasks,the effect of some current deep learning network models is not ade-quate.Based on the pain points of the existing models,a lightweight bird recognition method with more ad-vantages in recognition accuracy,computing performance,resource consumption and practicability is pro-posed.This method is based on the lightweight MobileNet model and blends the Convolutional Block Atten-tion Module(CB AM),which reduces the complexity of the model while enhancing the ability to extract lo-cal features.The comparative experiment with the mainstream model verifies the validity of the model.In addition,the model can be applied to hardware platforms such as robots and portable mobile phones,ex-panding the application scenarios,adding databases for auxiliary calculations,reducing additional costs,and having both portability and practicability.
关键词
MobileNet/注意力机制/Android平台/鸟类识别/图像识别Key words
MobileNet/attention mechanism/Android platform/bird recognition/image recognition引用本文复制引用
基金项目
国家自然科学基金(61501247)
国家自然科学基金(61703212)
国家自然科学基金(61802197)
国家自然科学基金(62071240)
出版年
2024
NETL
NSTL
万方数据