FasterRCNN - metrics

Hello everyone.

Does anyone know how we can compute the f1 score, specificity, sensitivity and ROC curve for predictions of pytorch FasterRCNN on a validation/test dataset?

Would appreciate any help!