Train on empty annotation (without objects)

[trueToastedCode]

I noticed that detecto does not train on images, where there are no object/s in the annotation.

My problem is, that my trained model detects objects with high confidence in completely different testing images, although there is nothing that is should detect.

I need to train on images with no objects in the annotation, so that it can learn not to detect something, when there is nothing.