How to do hard negative mining

[cygerts]

Hi,
I wonder how it would be possible to run hard negative mining as in SSD paper?
" Instead of using all the negative examples, we sort them using the highest confidence loss for each
default box and pick the top ones so that the ratio between the negatives and positives is
at most 3:1. "

[Eromera]

@cygi1989 I'm also interested, I tried to implement it from tf object detection framework but it doesnt seem to work correctly. Did you reach any conclusion? Thanks