-
Notifications
You must be signed in to change notification settings - Fork 12
Open
Description
说一下我的理解和问题:
- 做不了细分类的原因是,训练阶段的mini-batch是只获取IoU0.5以上的框,目的就是对已知instance建模。因此整个模型就没有考虑到unknown instance。只是在test的时候,或者说,相当于后处理方法,给目标一个分数来判断是否是已知或者未知,这也是为什么文中会说:“Furthermore, since background proposals usually overwhelm the mini-batch, we sample the same number of foreground and background proposals, enabling our model to recall unknown objects from the background class.”
- 我的问题是,我在 只找到正样本部分,没找到处理负样本的代码。这虽然会让正样本类会更近,但并不会让不同类的样本更远呀?
希望得到回复!谢谢
Metadata
Metadata
Assignees
Labels
No labels