How to manually add negative samples for training to improve precision

[leefionglee]

Hi All, I have successfully trained faster rcnn on my own dataset. It works with high recall but the precision is quite low with some false positives. I wonder if there is a way to explicitly add negative examples (i.e. as the background) to improve the precision? For example, I can add these false positive areas (from testing) as the negative samples to train a new model.

If yes, then how? thanks!!

[korkinof]

Anything on how to increase precision?

[ericromanenghi]

Could you find the way to add negative samples?

[nwestlake]

Set TRAIN.BG_THRESH_LO to 0.0?

[Supersak80]

@leefionglee an updates on this? I'm also having the same problem with high recall and low precision due to false positives.

@nw362 what is the intuition behind doing what you suggested?

[saiprabhakar]

I think backgrounds are considered negative samples. So you just need to find a way to add bounding boxes with overlap<threshold_overlap to your 'selective search file'.