color mode in face parsing module

[archershot]

I notice that the normailize parameters "mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225]" should be apply on PIL color mode (RGB). While in 'dataset', i found that images are loaded with Opencv color model(BGR). After i test both loading mode with pretrained weights, RGB mode performs as well as BGR mode. So, which mode should i choose?

[nttstar]

I think it's BGR input according to the code.