基于动态高效网络的花粉图像识别模型

Computer Engineering and Design(2023)

Cited 0|Views3
No score
Abstract
针对花粉图像清晰度不足、形状相近等特性所导致的识别精度低、人工识别低效的问题,提出一种基于动态高效网络的花粉图像识别模型.在ImageNet数据集上使用Noisy Student方法对EfficientNet进行预训练;将训练后的权重迁移到花粉识别模型中;引入动态学习率提升模型的识别精度.采用MixUp和GridMask的方法用于花粉图像数据增强.仿真结果与其它模型进行比较,验证了提出模型具有更好的分类效果.
More
Translated text
Key words
pollen image,deep learning,Efficientnet,data augmentation,noisy student,transfer learning,dynamic learning rate
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined