resnet18 ();

pretrainedURL ='//www.tianjin-qmedu.com/万博1manbetxsupportfiles/vision/data/deeplabv3plusResnet18CamVid.mat';pretrainedFolder = fullfile (tempdir,'pretrainedNetwork');pretrainedNetwork = fullfile (pretrainedFolder,'deeplabv3plusResnet18CamVid.mat');如果〜存在（pretrainedNetwork，'文件'mkdir (pretrainedFolder);disp (“预训练下载网（58 MB）......”);websave (pretrainedNetwork pretrainedURL);结束

imageURL =“http://web4.cs.ucl.ac.uk/staff/g.brostow/MotionSegRecData/files/701_StillsRaw_full.zip”;labelURL =“http://web4.cs.ucl.ac.uk/staff/g.brostow/MotionSegRecData/data/LabeledApproved_full.zip”;outputFolder =完整文件（TEMPDIR，“CamVid”);labelsZip = fullfile (outputFolder,'labels.zip');imagesZip = fullfile (outputFolder,“images.zip”);如果〜存在（labelsZip，'文件')| | ~存在(imagesZip'文件'mkdir (outputFolder) disp (“下载16层MB CamVid数据集的标签......”);websave（labelsZip，labelURL）;解压缩（labelsZip，完整文件（outputFolder，'标签'));disp (“正在下载557 MB的CamVid数据集图像……”);websave（imagesZip，IMAGEURL）;解压缩（imagesZip，完整文件（outputFolder，“图片”));结束

imgDir = fullfile (outputFolder,“图片”,'701_StillsRaw_full');imd = imageDatastore (imgDir);

I = readimage (imd, 1);I = histeq(我);imshow(我)

类= [“天空”“建筑”“极”“路”“路面”“树”“SignSymbol”“篱笆”“汽车”“行人”“自行车”];

labelIDs = camvidPixelLabelIDs ();

labelDir = fullfile (outputFolder,'标签');pxds = pixelLabelDatastore（labelDir，类，labelIDs）;

C = readimage (pxds, 1);提出= camvidColorMap;我= labeloverlay (C“ColorMap”，CMAP）;imshow（B）pixelLabelColorbar（CMAP，班）;

台= countEachLabel (pxds)

频率= tbl.PixelCount /笔(tbl.PixelCount);(1:numel(类)，频率)xticks(1:numel(类))xticklabel (tbl.Name) xtickangle(45) ylabel('频率')

[imdsTrain, imdsVal, imdsTest, pxdsTrain, pxdsVal, pxdsTest] = partitionCamVidData(imds,pxds);

numTrainingImages =元素个数(imdsTrain.Files)

numValImages = numel（imdsVal.Files）

numTestingImages = numel（imdsTest.Files）

%指定网络映像大小。这通常与训练图像的大小相同。imageSize = [720960 3];%指定类的数目。numClasses =元素个数(类);%创建DeepLab v3+。lgraph = deeplabv3plusLayers(imageSize, numClasses，“resnet18”);

imageFreq = tbl.PixelCount ./ tbl.ImagePixelCount;classWeights =中间值（imageFreq）./ imageFreq

pxLayer = pixelClassificationLayer（“名字”,'标签',“类”，tbl.Name，“ClassWeights”,classWeights);lgraph = replaceLayer (lgraph,“分类”,pxLayer);

定义验证数据。pximdsVal = pixelLabelImageDatastore (imdsVal pxdsVal);定义培训选项。选项= trainingOptions（“个”,...“LearnRateSchedule”,“分段”,...'LearnRateDropPeriod'10...“LearnRateDropFactor”，0.3，...“动量”,0.9,...“InitialLearnRate”，1E-3，...“L2Regularization”,0.005,...'ValidationData'，pximdsVal，...'MaxEpochs'30，...“MiniBatchSize”8...“洗牌”,“每个历元”,...'CheckpointPath'tempdir,...“VerboseFrequency”2,...“阴谋”,“训练进步”,...“ValidationPatience”4);

增量= imageDataAugmenter (“RandXReflection”，真正，...“RandXTranslation”-10年[10],“RandYTranslation”,-10年[10]);

pximds = pixelLabelImageDatastore (imdsTrain pxdsTrain,...“DataAugmentation”、增压器);

doTraining = FALSE;如果doTraining [net, info] = trainNetwork(pximds,lgraph,options);其他的data =负载(pretrainedNetwork);网= data.net;结束

I = readimage (imdsTest 35);C = semanticseg(I, net);

我= labeloverlay (C“Colormap”提出,“透明”,0.4);imshow (B) pixelLabelColorbar(提出、类);

expectedResult = readimage (pxdsTest 35);实际= uint8 (C);预期= uint8 (expectedResult);预计imshowpair(实际)

借据= jaccard (C, expectedResult);表(类、借据)

pxdsResults = semanticseg (imdsTest净,...“MiniBatchSize”4...“WriteLocation”tempdir,...“详细”，假）;

度量= evaluateSemanticSegmentation（pxdsResults，pxdsTest，“详细”，假）;

metrics.DataSetMetrics

metrics.ClassMetrics

功能labelIDs = camvidPixelLabelIDs ()％返回对应于每一类的标签的ID。%CamVid数据集有32个类。把他们分成11个班%原SegNet训练方法[1]。%% 11类是:%的“天空”“建筑”,“极”,“路”,“路面”、“树”、“SignSymbol”,“栅栏”、“汽车”、“行人”和“骑自行车的人”。%% CamVid像素标签id作为RGB颜色值提供。集团成% 11类，并将其作为m×3矩阵的单元数组返回。的每个RGB值旁边都列出了%原始的CamVid类名。请注意%，其他/Void类被排除在下面。labelIDs = {...%的“天空”[128 128 128;...%的“天空”]％ “建造”[000 128064;...％“桥”128 000 000;...％ “建造”064 192 000;...%的“墙”064 000 064;...%的“隧道”192 000 128;...%的“拱门”]％“极点”[192 192 128;...%”Column_Pole”000 000 064;...％ “交通拥挤”]%的道路[128064 128;...%的“路”128 000 192;...％ “LaneMkgsDriv”192 000 064;...%”LaneMkgsNonDriv”]%“路面”[100万192;...%“人行道”064 192 128;...％ “ParkingBlock”128 128 192;...%”RoadShoulder”]%的“树”[128128 000;...%的“树”192 192 000;...%”VegetationMisc”]%”SignSymbol”[192 128 128;...%”SignSymbol”128 128 064;...%”Misc_Text”000 064 064;...%”TrafficLight”]％“篱笆”[064 064 128;...％“篱笆”]%的“汽车”[064 000 128;...%的“汽车”064 128 192;...%”SUVPickupTruck”192 128 192;...%”Truck_Bus”192 064 128;...%“训练”128 064 064;...%”OtherMoving”]%“行人”[064 064 000;...%“行人”192 128 064;...%的“孩子”064 000 192;...％ “CartLuggagePram”064 128 064;...%的“动物”]%“自行车”[000 128192;...%“自行车”192 000 192;...％ “MotorcycleScooter”]};结束

功能一会pixelLabelColorbar(提出)向当前轴添加一个颜色栏。颜色栏被格式化了%以显示类名的颜色。颜色表（GCA，CMAP）%添加颜色栏到当前的数字。c = colorbar (“对等”，GCA）;％用途类名的刻度线。c。TickLabels =一会;numClasses =大小(提出,1);%中心标记标签。c。蜱虫= 1 / (numClasses * 2): 1 / numClasses: 1;清除标记。c。TickLength = 0;结束

功能提出= camvidColorMap ()%定义被CamVid数据集使用的颜色映射。cmap = [128,128,128]%的天空128 0 0％ 建造192 192 192%极128 64 128%的道路60 40 222%的人行道上128 128 0%的树192 128 128% SignSymbol64 64 128%的栅栏64 0 128%的车64 64 0%行人0 128 192%骑自行车];%在[0 1]之间正常化。cmap = cmap ./ 255;结束

功能[imdsTrain, imdsVal, imdsTest, pxdsTrain, pxdsVal, pxdsTest] = partitionCamVidData(imds,pxds)通过随机选择60%的数据进行训练，%分割CamVid数据。的% rest用于测试。设定初始随机状态，例如再现性。rng (0);numFiles =元素个数(imds.Files);shuffledIndices = randperm (numFiles);%使用60%的图像进行训练。numTrain = round(0.60 * numFiles);trainingIdx = shuffledIndices (1: numTrain);%使用20%的图像进行验证numVal = round(0.20 * numFiles);valIdx = shuffledIndices (numTrain + 1: numTrain + numVal);%剩余的用于测试。testIdx = shuffledIndices（numTrain + numVal + 1：结束）;创建用于培训和测试的图像数据存储。trainingImages = imds.Files (trainingIdx);valImages = imds.Files (valIdx);testImages = imds.Files (testIdx);imdsTrain = imageDatastore (trainingImages);imdsVal = imageDatastore (valImages);imdsTest = imageDatastore (testImages);提取类和标签id信息。类= pxds.ClassNames;labelIDs = camvidPixelLabelIDs ();创建用于培训和测试的像素标签数据存储。trainingLabels = pxds.Files (trainingIdx);valLabels = pxds.Files (valIdx);testLabels = pxds.Files (testIdx);pxdsTrain = pixelLabelDatastore(训练标签、类、标签);pxdsVal = pixelLabelDatastore(valLabels, classes, labelid);pxdsTest = pixelLabelDatastore(testlabel, classes, labelid);结束