为什么按照作者提供的代码训练之后,无法达到实验结果?
helix333111 opened this issue · 3 comments
helix333111 commented
以 resnet-v2 为backbone,split=0, epoch = 200 , 训练之后的结果远没有达到 mIoU 61.7( pascal 5i 1shot fold 0)?
作者提供的模型可以达到61.7 ,是不是epoch 为200太少了,还是batch size 为4 太小了?
我的训练结果只有 mIoU 接近 50
请教各位同学。
tianzhuotao commented
@helix333111
Hi, thanks for your interest.
You can take reference from this issue:
We only use 1 GPU to train our model on PASCAL, and different Pytorch versions may cause different results.
Thank you.
tianzhuotao commented
helix333111 commented
Thank you !!