使用MATLAB进行目标检测训练时出错

Question

clear;clc;
load('gTruth.mat');                               %标记的gTruth
trainingfiles = gTruth.DataSource.Source;         %图像位置
traininglabel = gTruth.LabelData;                 %图像标签

Combine the datastores.
imagefile = imageDatastore(trainingfiles);
imagelabel = boxLabelDatastore(traininglabel);
trainingData = combine(imgfiles, imglabel);

Load a preinitialized YOLO v2 object detection network.
inputImageSize = [227 227 3];
numClasses = 1;
anchorBoxes = [8,8; 12,12; 24 24];
network = 'resnet50';
featureLayer = 'activation_40_relu';
 
lgraph = fasterRCNNLayers(inputImageSize, numClasses, anchorBoxes,network,featureLayer); 


augmentedTrainingData = transform(trainingData,@augmentData);

augmentedData = cell(6,1);
for k = 1:6
    data = read(augmentedTrainingData);
    augmentedData{k} = insertShape(data{1},"rectangle",data{2},"LineWidth",3);
    reset(augmentedTrainingData);
end
figure
montage(augmentedData,BorderSize=10)

options = trainingOptions('sgdm',...
          'InitialLearnRate',0.001,...
          'Verbose',true,...
          'MiniBatchSize',10,...
          'MaxEpochs',20,...
          'Shuffle','never',...
          'VerboseFrequency',10,...
          'CheckpointPath',tempdir);

[detector,info] = trainFasterRCNNObjectDetector(augmentedTrainingData,lgraph,options);

错误使用 visiongpuROIMaxPoolingForward
Maximum variable size allowed on the device is exceeded.

出错 trainFasterRCNNObjectDetector (第 462 行)
[detector, info] = iTrainEndToEnd(trainingData, fastRCNN, options, params, executionSettings, imageInfo);

Answer 1

给变量分配的内存太大了，应该是你电脑配置问题，升级电脑或者调小申请内存大小（做机器学习的最好要高配一点的电脑，特别是用matlab的）