ImageClassification.Fit-IndexOutOfRangeException

Question

我已经使用Fit方法来训练我的模型，这会生成异常“索引在数组的边界之外”。

我创建了一个库.Net Core 3.1和一个UnitTesting（.Net Core 3.1），其中UnitTesting在我的库中将属性称为“ Training（）”。

此代码参考Tutorial: Automated visual inspection using transfer learning with the ML.NET Image Classification API中的示例

public void Training()
{
    MLContext mlc = new MLContext();

    try
    {
        IEnumerable<ImageData> images = LoadImagesFromDirectory(folder: AssetsRelativePath, useFolderNameAsLabel: true);
        IDataView imageData = mlc.Data.LoadFromEnumerable(images);
        IDataView shuffledData = mlc.Data.ShuffleRows(imageData);

        var preprocessingPipeline = mlc.Transforms.Conversion.MapValueToKey(inputColumnName: "Label", outputColumnName: "LabelAsKey")
            .Append(mlc.Transforms.LoadImages(outputColumnName: "Image", imageFolder: AssetsRelativePath, useImageType: false, inputColumnName: "ImagePath"));

        IDataView preProcessedData = preprocessingPipeline.Fit(shuffledData).Transform(shuffledData);

        TrainTestData trainSplit = mlc.Data.TrainTestSplit(data: preProcessedData, testFraction: 0.3);
        TrainTestData validationTestSplit = mlc.Data.TrainTestSplit(trainSplit.TestSet);

        IDataView trainSet = trainSplit.TrainSet;
        IDataView validationSet = validationTestSplit.TrainSet;
        IDataView testSet = validationTestSplit.TestSet;

        var trainingPipeline = mlc.Model.ImageClassification(
               featuresColumnName: "Image", 
               labelColumnName: "LabelAsKey", 
               arch: ImageClassificationEstimator.Architecture.ResnetV2101,
               epoch: 100, 
               batchSize: 10, 
               testOnTrainSet: false, 
               metricsCallback: (metrics) => Logger.Debug(metrics), 
               validationSet: validationSet,
               reuseTrainSetBottleneckCachedValues: true, 
               reuseValidationSetBottleneckCachedValues: true, 
               disableEarlyStopping: false
            )
            .Append(mlc.Transforms.Conversion.MapKeyToValue("PredictedLabel"));


        ITransformer trainedModel = trainingPipeline.Fit(trainSet);
    }
    catch{

    }
}

我收到例外：IndexOutOfRangeException

Answer 1

可能是您的某些图像无法加载？也许在训练之前尝试将trainSet保存到磁盘（作为文本或二进制文件），然后查看它是否继续抛出？

基本上，我怀疑数据本身已损坏，并且第一次扫描数据时，您立即发现了这一点。