基于Swin Transformer的动物图片分类模型研究

Research on animal picture classification model based on Swin Transformer

李旭昌 ¹肖鑫 ¹张曦心¹

扫码查看

作者信息

1. 西藏大学信息科学技术学院,拉萨 850000
折叠

摘要

高原地区的特殊环境给图像分类技术带来了额外的挑战.因此,提出一种基于Swin Transformer的图片分类模型.并与传统的卷积神经网络(CNN)和Vision Transformer(ViT)模型进行对比,通过网络爬虫、收集拍摄等方式搜集10000张图片数据用于模型的训练评估,该模型最终准确率为92.50%,精确率为91.80%,在相同数据集下表现优于CNN和ViT模型.

Abstract

The special environment of the plateau area brings additional challenges to image classification technology.There-fore,this paper proposes an image classification model based on Swin Transformer(Shift Window Transformer).Compared with the traditional Convolutional Neural Network(CNN)and Vision Transformer(ViT)models,10000 images were collected for training and evaluation by means of web crawlers,collection and shooting,and the final accuracy and precision of the model was 92.50%and 91.80%,respectively,which outperformed the CNN and ViT models under the same dataset.

关键词

Swin/Transformer/动物图片/分类模型/图像数据

Key words

Swin Transformer/animal images/classification model/image data

引用本文复制引用

出版年

2024

现代计算机

中大控股

现代计算机

影响因子：0.292

ISSN：1007-1423

段落导航