Recall and precision 00e+00

I am using attention unet model and dice loss to detect cerebral aneurysm in 2d brain images however I am getting absurd results for recall and precision per epoch (0.00e+00)

If your training labels are all negatives and there are no positives then you can have precision and recall values are 00e+00. To overcome you could try to downsampling the negative labels. Thank you