Detectron.pytorch
Detectron.pytorch copied to clipboard
Can this implementation handle images without objects?
Hello! I want to train a model with some samples that has no objects ( just background ). Can this implementation handle that?
可以
@13070151771 意思是只要把 roidb = filter_for_training(roidb) 去掉就可以吗? 我看代码,大多数都对无目标图可以处理,但是有少部分似乎就无法处理,改完之后loss经常突然增大