Error using vision.int​ernal.cnn.​utils.mean​StdForTarg​ets - object detection using Faster-RCNN

Hi all!

I'm trying to detect guns and apply bounding boxes around them by running faster RCNN code with VGG16 and I got this error before training, Did anybody known?

Error using vision.internal.cnn.utils.meanStdForTargets (line 9)
Unable to find any region proposals to use as positive training samples. Lower the first value of
PositiveOverlapRange to increase the number of positive region proposals.

Error in vision.internal.cnn.rcnnDatasetStatistics (line 34)
    [info.BoxRegressionMean, info.BoxRegressionStd] =
    vision.internal.cnn.utils.meanStdForTargets(out.targets);

Error in trainFasterRCNNObjectDetector>iCollectImageInfo (line 1637)
    imageInfo = vision.internal.cnn.rcnnDatasetStatistics(trainingData, rpnLayerGraph,
    imageInfoParams);

Error in trainFasterRCNNObjectDetector (line 418)
    [imageInfo,trainingData] = iCollectImageInfo(trainingData, fastRCNN, iRPNParamsEndToEnd(params),
    params);

Error in VGG16 (line 134)
    [detector, info] = trainFasterRCNNObjectDetector(trainingData,lgraph,options, ...

Answer 1

[Srivardhan Gadila]

PositiveOverlapRange is a two-element vector, used to specify the Bounding box overlap ratios for positive training samples. The vector contains values in the range [0,1]. Region proposals that overlap with ground truth bounding boxes within the specified range are used as positive training samples.

For more information, refer to the description under "PositiveOverlapRange" in the documentation of trainFastRCNNObjectDetector.

As the error message suggests, try lowering the first value. Even then if the error occurs then try contacting Technical Support.