Through the code test, I can clearly know the model training results, but the segmentation results has no corresponding evaluation index. So what I want to ask is whether it is possible to quantitatively compare the quality of the segmentation results?Thanks!