bubbliiiing/retinanet-pytorch

num_class=20

Rashfu opened this issue · 1 comments

Rashfu commented

您好，我想问一下为什么针对voc数据集，网络预测的分类不是21个？你后面算focal_loss不看背景吗

bubbliiiing commented

retinanet预测原理了解一下