一、环境配置与数据准备
1.1 环境要求
- MATLAB版本：R2021a及以上（需安装Deep Learning Toolbox）
- GPU支持：推荐NVIDIA CUDA兼容显卡（通过
gpuDevice验证）
1.2 数据组织结构
dataset/├── train/│ ├── cat/│ └── dog/└── validation/├── cat/└── dog/1.3 数据加载与预处理
% 创建图像数据存储imdsTrain=imageDatastore('dataset/train',...'IncludeSubfolders',true,...'LabelSource','foldernames');imdsValidation=imageDatastore('dataset/validation',...'IncludeSubfolders',true,...'LabelSource','foldernames');% 数据增强（随机旋转±20°，水平翻转）augmenter=imageDataAugmenter(...'RandRotation',[-20,20],...'RandXReflection',true);% 调整图像大小并增强augimdsTrain=augmentedImageDatastore([227227],imdsTrain,'DataAugmentation',augmenter);augimdsValidation=augmentedImageDatastore([227227],imdsValidation);二、模型构建策略
2.1 迁移学习（推荐方法）
% 加载预训练模型（AlexNet/ResNet-50/EfficientNet）net=alexnet;% 修改网络结构lgraph=layerGraph(net);newFCLayer=fullyConnectedLayer(2,'Name','fc_new','WeightLearnRateFactor',10);newOutputLayer=classificationLayer('Name','output_new');% 替换最后两层lgraph=replaceLayer(lgraph,'fc7',newFCLayer);lgraph=replaceLayer(lgraph,'ClassificationLayer_fc7',newOutputLayer);2.2 自定义CNN架构
layers=[imageInputLayer([2272273])% 卷积块1convolution2dLayer(3,32,'Padding','same')batchNormalizationLayer reluLayermaxPooling2dLayer(2,'Stride',2)% 卷积块2convolution2dLayer(3,64,'Padding','same')batchNormalizationLayer reluLayermaxPooling2dLayer(2,'Stride',2)% 全连接层fullyConnectedLayer(64)reluLayerdropoutLayer(0.5)% 输出层fullyConnectedLayer(2)softmaxLayer classificationLayer];三、模型训练与调优
3.1 训练参数配置
options=trainingOptions('adam',...'MaxEpochs',20,...'MiniBatchSize',64,...'InitialLearnRate',0.001,...'Shuffle','every-epoch',...'ValidationData',augimdsValidation,...'ValidationFrequency',30,...'Verbose',false,...'Plots','training-progress',...'ExecutionEnvironment','multi-gpu');% 支持多GPU加速3.2 模型训练
[netTrained,info]=trainNetwork(augimdsTrain,lgraph,options);3.3 性能评估
% 验证集预测YPred=classify(netTrained,augimdsValidation);YValidation=imdsValidation.Labels;% 计算准确率accuracy=mean(YPred==YValidation);fprintf('Validation Accuracy:%.2f%%',accuracy*100);% 混淆矩阵cm=confusionchart(YValidation,YPred);cm.Title='Confusion Matrix';cm.ColumnSummary='column-normalized';四、实战案例：花卉分类
5.1 数据集准备
下载并解压Oxford 102 Flowers数据集，按类别组织文件夹。
5.2 完整代码
% 加载数据[imdsTrain,imdsValidation]=loadFlowerDataset();% 数据增强augmenter=imageDataAugmenter('RandRotation',[-15,15]);augimdsTrain=augmentedImageDatastore([227227],imdsTrain,'DataAugmentation',augmenter);% 迁移学习net=alexnet;lgraph=layerGraph(net);layers=[lgraph.Layers(1:end-3)...% 移除最后3层fullyConnectedLayer(102,'WeightLearnRateFactor',10)...softmaxLayer...classificationLayer];% 训练配置options=trainingOptions('sgdm',...'MaxEpochs',15,...'MiniBatchSize',32,...'InitialLearnRate',0.001,...'ExecutionEnvironment','gpu');% 开始训练netTrained=trainNetwork(augimdsTrain,lgraph,options);% 评估模型YPred=classify(netTrained,imdsValidation);accuracy=mean(YPred==imdsValidation.Labels);五、模型部署
6.1 MATLAB实时推理
% 加载测试图像img=imread('test_flower.jpg');imgResized=imresize(img,[227227]);% 预测label=classify(netTrained,imgResized);imshow(img);title(sprintf('Predicted: %s (%.2f%%)',label,max(scores)*100));6.2 生成TFLite模型
converter=dlquantizer(netTrained,'Target','TensorFlow Lite');converter.Optimize=true;converter.Precision='int8';tfliteModel=convert(converter);save('flower_classifier.tflite','tfliteModel');十、参考
MathWorks官方文档：Deep Learning in MATLAB]ww2.mathworks.cn/help/deeplearning/
代码 运用深度学习模型实现图像的分类www.3dddown.com/csa/55199.html
AlexNet迁移学习示例：Image Category Classificationww2.mathworks.cn/help/deeplearning/ug/image-category-classification-using-deep-learning.html
