Distilling-Object-Detectors icon indicating copy to clipboard operation
Distilling-Object-Detectors copied to clipboard

关于mask_batch的问题

Open TangAL0203 opened this issue 6 years ago • 0 comments

repo主你好,我有下面3个问题,希望能得到你的解答:

  1. anchor_target_layer.py中,mask_per_gt = torch.sum(IOU_map[i][:,:,:,k]>max_iou_per_gt, dim = 2) 这里将HWK在K维度上求和,这个操作和论文中提到的 ”With F, we filter the IOU map to keep those larger then F locations and combine them with OR operation to get a W × H mask“ 不匹配呀?
  2. 另外,论文中的公式3,imitation loss weight balancing factor这个值取值为多少呢?代码中默认值是0.01
  3. repo主有没有试过当K等于1,HW上一个GT框可能就只有一个Loc能与之匹配,这种情况下做mimic还有效果吗?

TangAL0203 avatar Oct 22 '19 11:10 TangAL0203