search

Leotju / MGAN

Mask-Guided Attention Network for Occluded Pedestrian Detection. (ICCV'19)
128 stars 26 forks source link

Predictions from MGA layer used in loss function? #7

Open NehalD99 opened 4 years ago

NehalD99 commented 4 years ago

Hello,

I would like to ask whether the predictions from the MGA layer were directly used in the BCE Loss as stated in the paper? How did you modify the prediction output size(7x7) to match the visible box coarse annotation?

Thank you, Nehal

Leotju commented 4 years ago
  1. The loss is the same as stated in the paper.
  2. You could crop and resize the ground thuth.