reduce size of images for detection to reduce computations - is there a minimum image size to make the model work well?

[nibu (nibu147)]

I see that the detection model can be slow if an input image has a large size.
Are there any guidelines on an "optimal" image size to pass to the model?
Thanks!