Input image(s) for (top-down) inference - channel order BGR or RGB ?

[hfassold]

I want to use MMPose top-down inference (fn. mmpose.apis.inference_topdown), using an ndarray image loaded via OpenCV (so in BGR channel order) as input image. I am using the 'RTMPose-m' model for human skeleton estimation.

My question now is - do I have to convert the input image from BGR to RGB before handing it over to the MMPose inference routine ?
Is there are general guideline for MMPose - e.g. input images for inference have to be always in RGB format ?

[hfassold]

Hello Mister,

according to #1566 it seems that MMPose expects the input image in RGB.