Best score to compare binary mask

Hello there, im working on a project where the dataset given (MVTec) has some binary mask having the ground truth.

I made many model that give me back as a result another many binary mask. I would like to have a numeriacal value that rapresent the 'correctness' of the binary mask.

The mask could have many areas, of differenct shapes or sizes, like this : [https://imgur.com/a/3bLmU2w](https://imgur.com/a/3bLmU2w) .

My result are sometimes similar, sometime my result are inside the GT, sometimes …

binary computervision dataset hello project truth value