Overflow detected using TrainImagesClassifier

#1

Hello,

I’m working on a 1.6GB raster and running OTB’s TrainImagesClassifier in QGIS 3.6. And i would like to know how to avoid the below error. My vector data list has 86 features, and max training sample size per class is default ‘1000’ The process finishes and produces an output model, but am I losing anything with below issue, or should I ignore?

2019-03-27 21:42:17 (INFO): Sampling rates for image 1 : className requiredSamples totalSamples rate

5643 11984 0.470878

0 5643 2783412 0.00202737

1 5643 325897 0.0173153

2 5643 521180 0.0108274

4 5643 5643 1

5 5643 1465339 0.00385099

2019-03-27 21:42:17 (INFO): Sample selection…

2019-03-27 21:42:17 (INFO): Elevation management: setting default height above ellipsoid to 0 meters

2019-03-27 21:42:17 (INFO): Sampling strategy : set number of samples for each class

2019-03-27 21:42:17 (INFO): Sampling rates : className requiredSamples totalSamples rate

0 11984 0

0 5643 2783412 0.00202737

1 5643 325897 0.0173153

2 5643 521180 0.0108274

4 5643 5643 1

5 5643 1465339 0.00385099

5643 11984 0.470878 5643 0 0 [OVERFLOW]

2019-03-27 21:42:17 (WARNING): 1 case of overflow detected! (requested number of samples higher than total available samples)

Thanks for any help
Ross

0 Likes