Image labeler ground truth object doesn't provide valid output

Hi,

I have tried to retrieve training data using "objectDetectorTrainingData" from a ground truth object which was created using MATLAB 2017b image labeler (pixel-wise), and It doesn't give me an valid training data table/output.

However, it works fine when I used bounding boxes to select ROIs instead than pixel wise polygon. Could someone kindly help me to resolve this issue? I have attached my image, lebelling session and the ground truth object which I have received.

Many thanks in advance.

[Florian Morsch]

If you use the image labeler app you can change the type how your data should be saved (gTruth or table). Have you tried to save them as table and then use them to train?