Hi, when visualizing the boxes, they seem to be kind of unprecise. Are you sure that this cannot affect the model performance based on this dataset? On many images the people are not precisely surrounded by the boxes, some space is left. Or, some parts of the body are missing.

Thanks.