liminn / ICNet-pytorch

ICNet implemented by pytorch, for real-time semantic segmentation on high-resolution images, mIOU=71.0 on cityscapes, single inference time is 19ms, FPS is 52.6.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Ignore label question,标签处理问题

LEE-cm opened this issue · comments

commented

Cityscapes dataset, I changed the ignored label from the original -1 to 255 in original script and Correspondingly changed the loss function's ignore label, but meet errors.
为什么处理cityscapes数据集时把忽略的标签都设为-1,在计算损失函数时resize用的双线性差值不报错。当我把忽略标签改为255,计算损失函数就会报错,把插值方式改为近邻插值就没事了。

抱歉,由于工作原因很少有时间维护该git。
我猜测是双线性差值会改变像素值,比如出现254或253等lable中不存在的值。