Closed iperov closed 5 years ago
seriously ? your repo is wrong implemented
the cleaner sample, i.e. without attributes, the less loss should have.
you can show attribute_gt
and weight_attribute
in pfld/loss
file. you can see the more hard sample the more big weight have.
original paper:
your code:
1 / (x*256) if x > 0 else 1 for x in mat_ratio
you are decreasing weight_attribute if
the geometry and data imbalance functionalities
are present, but should be vice versa