CAPTAIN-WHU/DOTA_devkit

AP计算中的正样本统计

FL77N opened this issue · 1 comments

FL77N commented

您好,非常感谢您的工作!我有一个问题希望能得到您的解答:在计算task1的AP值时,计算正样本数量使用了npos = npos + sum(~difficult)这个来计算,请问为什么分类难度为difficult的样本不被计算在正样本之内呢?

This is the same as the PASCAL VOC metric, which is normal.