[question/potential bug] Scaling during average precision calculation

[TimoRST]

Hi there.
Recently I was rebuilding your metrics to evaluate models on custom data.
Whilst doing so, I stumbled upon the average precision calculation here: nuscenes-devkit/python-sdk/nuscenes/eval/detection/algo.py.

Subtracting the min_precision and then scaling by (1.0-min_precision) will scale down precisions below 1.0.
If you consider the case where all interpolated precisions are greater than min_precision, the result of mean(precision) is not equal to mean(max(precision-min_precision, 0.0))/(1.0-min_precision).
Shouldn't the function just be mean(max(precision-min_precision, 0.0))+min_precision?

Thanks in advance!

[whyekit-motional]

@TimoRST when calculating precision in the context of nuScenes, the evaluation protocol is only interested in operating points where the precision is > 10% (you can check out more details in Section 3.1 of https://arxiv.org/pdf/1903.11027)