Why is accuracy absent in training progress plot?

4 Ansichten (letzte 30 Tage)
Kevin Willeford
Kevin Willeford am 26 Sep. 2023
Beantwortet: Matt J am 26 Sep. 2023
I do not see training accuracy as illustrated in the docs. My code for training initialization is below and resultant plot attached.
% 6. specify training options
options = trainingOptions('sgdm', ...
BatchNormalizationStatistics = 'moving', ...
ExecutionEnvironment = 'auto', ...
GradientThreshold = 35, ...
InitialLearnRate = 0.0005, ...
LearnRateSchedule = 'piecewise', ...
LearnRateDropFactor = 0.99, ...
LearnRateDropPeriod = 1, ...
MaxEpochs = 20, ...
MiniBatchSize = 4, ...
Momentum = 0.9, ...
OutputNetwork = 'best-validation-loss', ...
Plots = 'training-progress', ...
ResetInputNormalization = false, ...
ValidationData = validationDS, ...
ValidationFrequency = 25, ...
VerboseFrequency = 5);
[trainedModel, info] = trainSOLOV2(trainingDS, preTrainedModel, options, ...
FreezeSubNetwork = "backbone", ExperimentMonitor = experiments.Monitor);

Antworten (1)

Matt J
Matt J am 26 Sep. 2023
Probably because the examples in the docs are for trainNetwork, not trainSOLOV2.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by