Calculate mAP for Segmentation task

[Mirzyaaliii]

I am working on a segmentation task (specifically, binary segmentation) and I want to calculate mAP between mAP. However, I am not sure how to calculate mAP between 2 masks (i.e., ground truth and prediction), as shown example in the below image.

Also, we cannot use the COCOeval script for that as it requires a score given by the model. And segmentation models like U-Net and FPN only generate masks without predicting the score for each masked region. (Please let me know if I am missing anything for this part.)

It would be helpful if anyone can point out to useful implementation/code (to calculate mAP for segmentation) or give some insights.