Hi, I have a classified image (Maximum Likelihood) that I want to run an accuracy assessment on, however I am unsure of the correct steps in ENVI. I have done it before in Erdas Imagine where a set of random points is generated and then you determine the "true" class of each point, which then results in an overall accuracy % for the classification. I am using 10m Sentinel imagery to identify the "true" classes.

I have tried the "Generate Random Sample Using Ground Truth Image" tool using my classified MLC image as input, used Stratified Random and input the amount of pixels I require for each category. So I now have ROIs for each classification category, but I am unsure of the next step to do something similar as in Erdas Imagine.

Thanks.