Evaluation Metrics for k-means Image Segmentation

Hi everyone!

When I used CNNs to segment an image base, I had available the evaluateSemanticSegmentation( ) function and its ClassMetrics and ImageMetrics objects, which provided me with metrics such as IoU, Accuracy, and so on.

However, when using k-means clustering, there is no equivalent function to obtain the desired metrics. Since I'm quite new to Matlab, I still am not able to write code to get these metrics. Basically there is a folder with test images and another one with their associated labels, and what I want is to get the IoUs and Accuracies between what was segmented by k-means and what is determined by the label for each image.

Could someone help me with this? =(

I am very grateful!

[engineer]

To calculate IoU;

you can use the following function;

bboxOverlapRatio