How does mAP metric deal with unlabeled objects in the validation set

Question

I am working on an object detection problem for which there are several objects in the images that all objects may not be labelled. I think that's the reason I am getting a low mAP score. However, when I evaluate the model on a test set, it detects the objects correctly, but may not find all of them in an image.

Let's say the image below is in the validation set, and only the left cat has been labelled.

When the trained model wants to evaluate this image, it detects both cats.

My question is: How is the mAP being calculated for this image? Is the detected cat without a label (the right one) a false positive?

Thanks.