zonasw / unet-nested-multiple-classification Goto Github PK

老哥，能给下数据处理那块的代码吗，将json转成图片，再获取全局类别，将图片转成uint8的格式

你好，用你的代码和你的数据复现，损失韩式很大，验证集才６０几，是否有问题？

运行时报这种错，是需要训练集中各个图像的size必须一致吗？

inference的结果里能不能加上map能指标

for masks, I make a multi-classification for 10 classes, I set the value of background as "0" for gray image, but in the testing result, I can always see the original image for the background, which means the background have not been classified. could you bring me some suggestions? should I begin background in value "1" in masks?

When I training the UNet with original settings and dataset , nothing is classified, as is shown in the image

INFO: Using device cuda
INFO: Network:
NestedUNet model
3 input channels
3 output channels (classes)
Bilinear upscaling
INFO: Creating dataset with 20 examples
training stops after this. can anyone help

这位大哥，我在用你框架的时候，用的是9类的（算上背景）分割，效果感觉很糟，是不是我数据太少了，才400张左右==

How to deal with it。

请问这个代码训练的时候不需要添加预训练模型么？