Ground-truth bounding boxes included in RPN proposals?

Hi,

I am training a Faster-RCNN-like architecture and wanted to log the detection losses on the validation dataset. To obtain these losses, I passed in the targets of the validation data to the forward() method of GeneralizedRCNN. When investigating the results, however, I noticed that each ground-truth bounding box had a near-exact match with a predicted bounding box. When passing targets=None, these bounding boxes were gone. Is this the intended behavior? If so, what is the rationale behind it?

I figured that this line is where the predicted and ground-truth boxes get mixed:

https://github.com/pytorch/vision/blob/095cabb76cdcd4763bad629481d189b91e3df42c/torchvision/models/detection/roi_heads.py#L646

It seems to have something to do with balancing positive/negative sampling. Yet, I don't see how a model would improve from boxes that it did not generate itself.

Read torchvision/models/detection/roi_heads.py around line 646 and the GeneralizedRCNN.forward path. Compare proposal generation when targets are supplied versus targets=None, then determine whether the near-matching ground-truth boxes are intentional and what rationale or behavior should be documented or changed.