Problem with confusion matrix after CNN application
So, I am having trouble with the execution of the confusion matrix. I prepared the rule set in the following way: First I do a chessboard segmentation, apply values for ground truth from the thematic layer for checkup and then transform that into the samples within the eCog. After that, I remove the classification and classify using training data, generate samples for CNN, train the network and apply it. Now I try to calculate the confusion matrix and get only null values and the sample values I told eCog to remember are gone. Any idea what is going on?
Another thing, why when I try to export the results as raster and want to have the value of classification to match the class (1, 2,3, 4, 5 - names of classes) it pushes out random numbers and does not use the numbers of class? I would love to know if there is a way to do so, it will help a lot and void excess processing.
Was this article helpful?