Bug: Running the Train Deep Learning Model Tool with Less than Ten Training Samples (Chips

Description

When running the Train Deep Learning Model tool, the error "ZeroDivisionError: division by zero" will be returned if there are less than ten training samples (chips) provided for each class of features being classified.

Cause

For example, let's say there are four training samples for one Class of features, and we use ten in the %Validation parameter. In this case, the number of samples for validation will be 0.4, which is not valid. The % value must be a whole number.

Workaround

The workaround for the issue is to make sure you have collected at least ten training samples for each category of feature to classify from your imagery.

Article ID: 000034256

Software: