How can I evaluate different detection model (means calculate mAP etc) ?

[BBBBBing]

Hello everyone

I bought a hailo raspberry pi development-kit and want to try different AI model for people detection.
I know that a lot of pre-trained detection models is evaluated already and I can find the benchmark of mAPs and fps.

My question is that how can I evaluate these models by my own dataset (annotated with vott format)?
Just like calculate mAPs .
It will be so appreciate if there is a python script to evaluate different models.

Thank you very much.