How to count number of false positives and false negatives in object detection I've got a rather simple question for you, however to which I can't find a proper answer. Let's have a simple setting which is visualized in the image below. We've got one object of class A in the image and our model predicts the occurrence of said class A with its position incorrectly estimated.
My question is whether in this example, we've got one false positive and that's all or it's not only a FP but simultaneously a false negative, since there's no prediction associated with ground truth annotation.

I'm trying to compute number of TP,FP,FN in my current dataset (mitosis count) to further compute F1 score, and I'm not sure whether I'm counting the number of occurrences of FP and FN correctly or whether annotations are, in terms of one class, only classified as FN when there's fewer number of predictions made than the number of ground truth annotations
Thank you, in advance
 A: This counts as both a false negative (missing the real object), and a false positive (predicting one where there is nothing).
To calculate these, you have to calculate overlapping area, to pair up the predictions and the true instances. A common threshold is to count everything as correct when the overlap is larger than 50% for example.
Even with an established overlap threshold, there are still some nuances to take into account when you have multiple predictions for the single true object, you should only count one of them as correct.
Find more info with the search terms IoU (intersection over union) and mAP (mean average precision), for example https://github.com/rafaelpadilla/Object-Detection-Metrics#metrics.
A: Probably FP and FN but I would say the way of measuring occurences in obejct detection is highly varibale so you have to go with what fits best for your case. If there is no overlap then your model predicted an object where there was none (FP) and it missed the real object (FN)
