| 基于迁移学习的家猪图像识别研究 |
| |
| 引用本文： | 谢碧森,段 清,刘俊晖,廖 赟,张 逸.基于迁移学习的家猪图像识别研究[J].教育技术导刊,2020,19(7):36-40. |
| |
| 作者姓名： | 谢碧森 段 清 刘俊晖 廖 赟 张 逸 |
| |
| 作者单位： | 云南大学 软件学院，云南 昆明 650504 |
| |
| 摘 要： | 为实现家猪图像识别并提高识别准确率，提出一种基于迁移学习的家猪图像识别方法。首先对现有数据集进行数据增强，然后迁移 VGG16 模型并对其进行微调，从而更好地提取图像特征并缩短网络训练时间。采用自归一化神经网络解决了梯度消失和梯度爆炸问题，在网络构造时使用全局平均池化代替全连接层，以达到降低模型过拟合的效果。实验对比结果表明，该方法分类效果较好，准确率达到了 84%，召回率和 F1 值分别提升至 0.8、0.82，各项指标相比基础模型均有所提升。
|
| 关 键 词： | 家猪图像识别 VGG16 迁移学习 自归一化 全局平均池化 |
| 收稿时间： | 2019-10-14 |
Image Recognition of Domestic Pigs Based on Transfer Learning |
| |
| XIE Bi-sen,DUAN Qing,LIU Jun-hui,LIAO Yun,ZHANG Yi.Image Recognition of Domestic Pigs Based on Transfer Learning[J].Introduction of Educational Technology,2020,19(7):36-40. |
| |
| Authors: | XIE Bi-sen DUAN Qing LIU Jun-hui LIAO Yun ZHANG Yi |
| |
| Institution: | School of Software，Yunnan University，Kunming 650504，China |
| |
| Abstract: | In order to realize and improve the recognition accuracy of domestic pig image，an image recognition method based on transfer learning is proposed. Firstly，the existing data set is enhanced，then VGG16 model is mitigated and fine-tuned to better extract image features and shorten the training time of the network. The problem of gradient disappearance and gradient explosion is solved by using the self-normalized neural network. In network construction，global average pooling is used instead of full connection layer to reduce overfitting. The experimental comparison results show that the classification effect of this method is good，with an accuracy of 84%，the recall rate and F1 value are improved to 0.8 and 0.82 respectively，and all indicators are improved compared with the basic
model. |
| |
| Keywords: | domestic pigs image identification VGG16 migration learning self-normalized global average pooling |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
