Alagirl

关于判别器的输出

1

想请问一下在训练得到confidence map过程中判别器的输出map大概是怎样的呢，我现在一直找不到分割网络和判别网路的训练平衡点，我做的肝脏病灶分割，分割网络预先训练使其达到一定的分割效果才加的判别器，判别器loss会降到很低，加上判别器分割精度会下降，请问您对我的训练有什么建议吗

I can not find the class 'Bar' in the code. Anyone could help me please!

2

in the train.py ,code import Bar and display train process.But I can not find the class define in the utils. Please help me. from utils import Bar, Logger, AverageMeter, accuracy,...

做肝脏分割一阶段粗分割得到的dice只有65%算正常吗?好像太低了吧

7

做肝脏分割一阶段粗分割得到的dice只有65%算正常吗?好像太低了吧,用Unet3d也可以很快的训练达到同样的精度,想问一下你们的实验也是这样吗,还是我中间有错误

label image resize

1

when I train the cityspace dataset,I need to resize image into (512,256,3).But the label value changed after I use cv2.resize function with nearest mode.How can I resize the label image...