PaddleDetection
PaddleDetection copied to clipboard
ppyoloer about the code question
问题确认 Search before asking
- [X] 我已经搜索过问题,但是没有找到解答。I have searched the question and found no related answer.
请提出你的问题 Please ask your question
在旋转目标检测的任务中,ppdet/modeling/heads/ppyoloe_r_head.py里面第371行-372行,对mask_positive增加维度之后又进行了tile的操作,相当于在最后一个维度复制了angle_max+1次,这个操作是为什么呢?不是太理解
Is there anyone can answer my question?
可以看下Density-aware Focal Loss 。这个操作是为了将每个正样本(即与真实框匹配或满足某种正样本条件的预测框)与一系列离散化的角度类别相关联,以便能够计算每个正样本预测框在所有角度类别上的预测损失
The issue has no response for a long time and will be closed. You can reopen or new another issue if are still confused.
From Bot