outputFolder = fullfile (tempdir,“Pandaset”）;lidarURL = [“https://ssd.mathworks.com/supportfiles/lidar/data/”.．.“Pandaset_LidarData.tar.gz”];helperDownloadPandasetData (outputFolder lidarURL);

modelName =“tiny-complex-yolov4-pandaset”；mdl = downloadPretrainedComplexYOLOv4 (modelName);网= mdl.net;

路径= fullfile (outputFolder,激光雷达的）;lidarData = fileDatastore(路径,“ReadFcn”@ (x) pcread (x));

gtPath = fullfile (outputFolder,“长方体”，“PandaSetLidarGroundTruth.mat”）;data =负载(gtPath“lidarGtLabels”）;标签= timetable2table (data.lidarGtLabels);boxLabels =标签(:,2:结束);

图ptCld = read(lidarData);ax = pcshow (ptCld.Location);集(ax,“XLim”50 [-50],“YLim”, 40 [-40]);变焦(ax, 2.5);轴从；

xMin = -25.0;xMax = 25.0;yMin = 0.0;yMax = 50.0;zMin = -7.0;zMax = 15.0;

bevHeight = 608;bevWidth = 608;

gridW = (yMax - yMin)/bevWidth;gridH = (xMax - xMin)/bevHeight;

yMin gridParams = {{xMin, xMax, yMax, zMin, zMax}, {bevWidth, bevHeight}, {gridW, gridH}};

writefile = true;如果writefile transformPCtoBEV (lidarData boxLabels、gridParams outputFolder);结束

dataPath公司= fullfile (outputFolder,“BEVImages”）;imd = imageDatastore (dataPath公司);

labelPath = fullfile (outputFolder,“长方体”，“BEVGroundTruthLabels.mat”）;负载(labelPath“processedLabels”）;建筑物= boxLabelDatastore (processedLabels);

(imd,建筑物)= removeEmptyData (imd,建筑物);

rng (0);shuffledIndices = randperm(大小(imds.Files, 1));idx = floor(0.6 * length(shuffledIndices));

imdsTrain =子集(imd, shuffledIndices (1: idx));imdsTest =子集(imd, shuffledIndices (idx + 1:结束);

bldsTrain =子集(建筑物,shuffledIndices (1: idx));bldsTest =子集(建筑物,shuffledIndices (idx + 1:结束);

trainData =结合(imdsTrain bldsTrain);testData =结合(imdsTest bldsTest);

validateInputDataComplexYOLOv4 (trainData);validateInputDataComplexYOLOv4 (testData);

networkInputSize = [608 608 3];isRotRect = true;preprocessedTrainingData =变换(trainData @(数据)preprocessData(数据、networkInputSize isRotRect));

data =阅读(preprocessedTrainingData);

我={1 1}数据;bbox ={1,2}数据;标签={1,3}数据;helperDisplayBoxes (bbox,标签);

重置(preprocessedTrainingData);

rng(0) isRotRect = false;trainingDataForEstimation =变换(trainData @(数据)preprocessData(数据、networkInputSize isRotRect));numAnchors = 6;[anchorBoxes, meanIoU] = estimateAnchorBoxes (trainingDataForEstimation numAnchors)

指定用于培训的类名。。一会= {“汽车”“卡车”“Pedestrain”};[净,networkOutputs anchorBoxes] = configureComplexYOLOv4(网,一会,anchorBoxes modelName);

maxEpochs = 90;miniBatchSize = 8;learningRate = 0.001;warmupPeriod = 1000;l2Regularization = 0.001;penaltyThreshold = 0.5;速度= [];

如果canUseParallelPool dispatchInBackground = true;其他的dispatchInBackground = false;结束mbqTrain = minibatchqueue (preprocessedTrainingData 2.．.“MiniBatchSize”miniBatchSize,.．.“MiniBatchFcn”@(图片,盒子,标签)createBatchData(图片,盒子,标签,类名),.．.“MiniBatchFormat”，[“SSCB”，＂＂),.．.“DispatchInBackground”dispatchInBackground,.．.“OutputCast”，[＂＂，“替身”]);

doTraining = false;如果doTraining迭代= 0;为学习率和小批丢失创建子图。无花果=图;[lossPlotter, learningRatePlotter] = configureTrainingProgressPlotter(图);%自定义训练循环。为epoch = 1:maxEpochs reset(mbqTrain);洗牌(mbqTrain);而(hasdata(mbqTrain))迭代=迭代+ 1;[XTrain, YTrain] =下一个(mbqTrain);使用dlfeval和% modelGradients函数。(渐变、州lossInfo) = dlfeval (@modelGradients,净、XTrain YTrain, anchorBoxes, penaltyThreshold, networkOutputs);%应用L2正则化。gradient = dlupdate(@(g,w) g + l2Regularization*w, gradient, net.Learnables);确定当前学习率值。currentLR = piecewiseLearningRateWithWarmup(迭代,时代、learningRate warmupPeriod, maxEpochs);使用SGDM优化器更新网络可学习参数。(净、速度)= sgdmupdate(净、渐变速度,currentLR);更新dlnetwork的状态参数。网=状态;%显示进展。如果国防部(迭代,10)= = 1 displayLossInfo(时代、迭代、currentLR lossInfo);结束用新点更新训练图。updatePlots (lossPlotter learningRatePlotter、迭代、currentLR lossInfo.totalLoss);结束结束其他的网= mdl.net;anchorBoxes = mdl.anchorBoxes;结束

创建一个表来保存返回的边界框、分数和标签%的探测器。结果=表(“大小”[0 3],.．.“VariableTypes”, {“细胞”，“细胞”，“细胞”}，.．.“VariableNames”, {“盒子”，“分数”，“标签”});在测试集中的图像上运行检测器并收集结果。重置(testData)而hasdata (testData)读取数据存储并获取映像。data =阅读(testData);形象={1 1}数据;运行检测器。executionEnvironment =“汽车”；(bboxes、分数、标签)= detectComplexYOLOv4(净、图像、anchorBoxes一会,executionEnvironment);收集结果。台=表({bboxes},{},{}标签,“VariableNames”, {“盒子”，“分数”，“标签”});结果=[结果;(资源);结束使用平均精度度量来评估目标检测器。metrics = evaluateDetectionAOS(结果，testData)

读取数据存储。reset(testData) data = read(testData);%获取图像。我={1 1}数据;运行检测器。executionEnvironment =“汽车”；(bboxes、分数、标签)= detectComplexYOLOv4 (anchorBoxes,净,我一会,executionEnvironment);%显示输出。图helperDisplayBoxes(我bboxes标签);

lidarTestData =子集(lidarData shuffledIndices (idx + 1:结束);ptCld =阅读(lidarTestData);[ptCldOut, bboxCuboid] = transferbboxToPointCloud (bboxes、gridParams ptCld);helperDisplayBoxes (ptCldOut、bboxCuboid、标签);

函数[gradients,state,info] = modelGradients(net,XTrain,YTrain,anchors,penaltyThreshold,networkOutputs) inputImageSize = size(XTrain,1:2);在CPU中收集基本事实以进行后处理。。YTrain =收集(extractdata (YTrain));从网络中提取预测。。[YPredCell、州]= complexYOLOv4Forward(网络,XTrain networkOutputs,锚);收集CPU中的激活以进行后处理并提取dlarray数据。集合预测= cellfun(@ gather,YPredCell(:，1:8)，“UniformOutput”、假);@ extractdata, gatheredforecasts，“UniformOutput”、假);%将预测从网格单元格坐标转换为框坐标。tiledAnchors = generateTiledAnchorsComplexYolov4 (gatheredPredictions(:, 2:5),锚);gatheredPredictions (:, 2:5) = applyAnchorBoxOffsetsComplexYolov4 (tiledAnchors, gatheredPredictions (:, 2:5) inputImageSize);从地面真实数据生成预测目标。。[boxTarget, objectnessTarget classTarget、objectMaskTarget boxErrorScale] = generateComplexYOLOv4Targets (gatheredPredictions、YTrain inputImageSize,锚,penaltyThreshold);%计算损失。boxLoss = bboxOffsetLoss(YPredCell(:，[2 3 9 10 6 7])，boxTarget,objectMaskTarget,boxErrorScale);objLoss = objectnessLoss (YPredCell (: 1), objectnessTarget, objectMaskTarget);clsLoss = classConfidenceLoss (YPredCell (:, 8), classTarget, objectMaskTarget);totalLoss = boxLoss + objLoss + clsLoss;信息。boxLoss = boxLoss;信息。objLoss = objLoss;信息。clsLoss = clsLoss; info.totalLoss = totalLoss;计算与损失相关的可学习变量的梯度。。梯度= dlgradient (totalLoss net.Learnables);结束

函数boxLoss = bboxOffsetLoss(boxPredCell,boxDeltaTarget,boxMaskTarget,boxErrorScaleTarget) lossX = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，1)，boxDeltaTarget(:，1)，boxMaskTarget(:，1)，boxErrorScaleTarget));lossY = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，2)，boxDeltaTarget(:，2)，boxMaskTarget(:，1)，boxErrorScaleTarget));lossW = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，3)，boxDeltaTarget(:，3)，boxMaskTarget(:，1)，boxErrorScaleTarget));lossH = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，4)，boxDeltaTarget(:，4)，boxMaskTarget(:，1)，boxErrorScaleTarget));lossYaw1 = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，5)，boxDeltaTarget(:，5)，boxMaskTarget(:，1)，boxErrorScaleTarget));lossYaw2 = sum(cellfun(@(a,b,c,d) mse(a.*c.*d,b.*c.*d)，boxPredCell(:，6)，boxDeltaTarget(:，6)，boxMaskTarget(:，1)，boxErrorScaleTarget));boxLoss = lossX +损耗+ lossW + lossH + lossYaw1 + lossYaw2;结束

函数clsLoss = classConfidenceLoss(classPredCell,classTarget,boxMaskTarget) clsLoss = sum(cellfun(@(a,b,c) crossentropy(a.*c,b.*c，“TargetCategories”，“独立”)、classPredCell classTarget boxMaskTarget (:, 3)));结束

函数objLoss = objectnessLoss(objectnessPredCell,objectnessDeltaTarget,boxMaskTarget) objLoss = sum(cellfun(@(a,b,c) crossentropy(a.*c,b.*c，“TargetCategories”，“独立”)、objectnessPredCell objectnessDeltaTarget boxMaskTarget (:, 2)));结束

函数data = preprocessData(数据、targetSize isRotRect)调整图像的大小，并将像素缩放到0和1之间。同时规模%对应的边界框。为ii = 1:size(data,1) I = data{ii,1};imgSize =大小(I);将单通道的输入图像转换为三个通道。如果numel(imgSize) < 3 I = repmat(I,1,1,3);结束bboxes = {ii, 2}数据;我= im2single (imresize(我targetSize (1:2)));规模= targetSize(1:2)。/ imgSize (1:2);bboxes = bboxresize (bboxes、规模);如果~isRotRect bboxes = bboxes(:，1:4);结束data(ii, 1:2) = {I,bboxes};结束结束函数[XTrain, YTrain] = createBatchData(数据、groundTruthBoxes groundTruthClasses一会)返回沿XTrain和批处理维度组合的图像%归一化边界框与YTrain中的classid连接。沿批处理尺寸连接图像。。XTrain =猫(4、数据{:1});从类名中获取类id。一会= repmat({分类(类名)},大小(groundTruthClasses));[~, classIndices] = cellfun (@ (a, b) ismember (a、b), groundTruthClasses,一会,“UniformOutput”、假);将标签索引和训练图像大小附加到缩放的边界框。%，并创建响应的单个单元格数组。combinedResponses = cellfun (@ (bbox, classid) [bbox, classid], groundTruthBoxes, classIndices,“UniformOutput”、假);len = max (cellfun (@ (x)大小(x, 1), combinedResponses));paddedBBoxes = cellfun(@(v) padarray(v，[len-size(v,1)，0]，0，“职位”)、combinedResponses“UniformOutput”、假);YTrain =猫(4,paddedBBoxes {: 1});结束

函数currentLR = piecewiseLearningRateWithWarmup(迭代,时代、learningRate warmupPeriod, numEpochs)%分段elearningratewithwarmup函数计算当前基于迭代次数的学习率%。持续的warmUpEpoch;如果迭代< = warmupPeriod增加热身期间迭代次数的学习率。currentLR = learningRate *((迭代/ warmupPeriod) ^ 4);warmUpEpoch =时代;elseif迭代>= warmupPeriod && epoch < warmUpEpoch+floor(0.6*(numepoch -warmUpEpoch))热身期结束后，如果剩余课时数小于60%，则保持学习速率不变。currentLR = learningRate;elseifepoch >= warmUpEpoch+floor(0.6*(numepoch -warmUpEpoch)) && epoch < warmUpEpoch+floor(0.9*(numepoch -warmUpEpoch))%如果剩余的代数大于但小于60%%超过90%，则学习率乘以0.1。currentLR = learningRate * 0.1;其他的如果还剩下90%以上的时间，就把所学的内容加倍。%率0.01。currentLR = learningRate * 0.01;结束结束

函数[lossPlotter, learningRatePlotter] = configureTrainingProgressPlotter (f)创建子图来显示损失和学习速率。图(f);clf次要情节(2,1,1);ylabel (学习速率的）;包含(“迭代”）;learningRatePlotter = animatedline;次要情节(2,1,2);ylabel (“全损”）;包含(“迭代”）;lossPlotter = animatedline;结束函数displayLossInfo(时代、迭代,currentLR lossInfo)显示每次迭代的丢失信息。disp (”时代:“+时代+" |迭代:"+迭代+“|学习率:”+ currentLR +.．." |全损:"+双(收集(extractdata (lossInfo.totalLoss))) +.．." |盒子丢失:"+双(收集(extractdata (lossInfo.boxLoss))) +.．." |对象丢失:"+双(收集(extractdata (lossInfo.objLoss))) +.．." |班级损失:"+双(收集(extractdata (lossInfo.clsLoss))));结束函数updatePlots (lossPlotter learningRatePlotter、迭代、currentLR totalLoss)更新损耗和学习率图。addpoints (lossPlotter、迭代、双(extractdata(收集(totalLoss))));addpoints (learningRatePlotter迭代,currentLR);drawnow结束函数helperDisplayBoxes (obj bboxes、标签)显示图像和点云上方的方框。。数字如果~ isa (obj,“pointCloud”) imshow(obj) shape =“矩形”；其他的pcshow (obj.Location);形状=“长方体”；结束showShape(形状,bboxes(标签= =“汽车”:).．.“颜色”，“绿色”，“线宽”, 0.5);在；showShape(形状,bboxes(标签= =“卡车”:).．.“颜色”，“红色”，“线宽”, 0.5);showShape(形状,bboxes(标签= =“Pedestrain”:).．.“颜色”，“黄色”，“线宽”, 0.5);持有从；结束函数helperDownloadPandasetData (outputFolder lidarURL)从给定的URL下载数据集到输出文件夹。。lidarDataTarFile = fullfile (outputFolder,“Pandaset_LidarData.tar.gz”）;如果~存在(lidarDataTarFile“文件”mkdir (outputFolder);disp (正在下载PandaSet Lidar驾驶数据(5.2 GB)…）;websave (lidarDataTarFile lidarURL);解压(lidarDataTarFile outputFolder);结束%解压文件。如果(~存在(fullfile (outputFolder激光雷达的)，“dir”）).．.& & (~ (fullfile (outputFolder,存在“长方体”)，“dir”)解压(lidarDataTarFile outputFolder);结束结束