基于迁移学习的垃圾分类平台设计与实现  

Design and Implementation of Garbage Classification Platform Based on Transfer Learning

作　　者：陈胜娣 Chen Shengdi(College of Computer Engineering,Maoming Polytechnic,Maoming,Guangdong 525027,China)

机构地区：[1]茂名职业技术学院计算机工程系,广东茂名525027

出　　处：《计算机时代》2025年第3期36-39,44,共5页Computer Era

基　　金：茂名市科技计划项目(230413104559623)。

摘　　要：为提高垃圾分类的效率和准确性,提供便民化的垃圾分类功能,设计并实现了基于迁移学习的垃圾分类平台。构建基于ResNet-18的特征提取骨干网络,引入ReLU激活函数和Dropout层以优化模型结构,并结合预训练权重进行迁移学习,最终该模型在自建垃圾分类数据集上取得92.24%的识别准确率。基于Django框架开发垃圾分类Web平台,支持用户通过网页上传垃圾图片,平台可即时预测并展示垃圾类别。To improve the efficiency and accuracy of garbage classification,this paper designs and implements a garbage classification platform based on transfer learning.Build a feature extraction backbone network based on ResNet-18,with the introduction of ReLU activation functions and Dropout layers to optimize the model structure,use pre-training weights for transfer learning,and finally the model achieved a recognition accuracy of 92.24%on a self-built garbage classification dataset.Additionally,develop a garbage classification web platform based on the Django framework,enabling users to upload images of garbage through the webpages,and the platform will instantly predict and display the garbage category.

关 键 词：垃圾分类 迁移学习 ResNet-18 DJANGO 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]