zhenqi-he / transnuseg Goto Github PK

Hello,

I tried to reproduce it using the histology dataset mentioned in the paper and the code on github, but it looks like there may be overfitting from the drop in the loss. All the parameters are set the same as in the paper, but the validation loss is basically not dropping and even increasing after 30 epochs. I've tried many times but still no solution, maybe there are some settings I'm not aware of? Also, the paper mentions "Rsd is initially set to 1 with a 0.3 decrease for every 10 epochs until it reaches 0.4", but the code on github sets ratio_d to 0 after 40 epochs, is that where the problem lies? Hope to get your help on this!

您好 感谢您写出的论文 在复现实验中我有几个疑惑希望您能解答 是关于模型最终输出的结果的 模型在训练结束后 保存了模型并输出了相应的指标 但是我发现在文章中对于数据集上的测评指标 在模型最后保存的时候只输出了一个 对于其他的指标如何计算 在代码中并没有指出 请问文章中对应指标是如何计算出来的

感谢您的论文，在训练阶段，模型的loss和对应图像可以存到日志当中，但模型训练的权重无法保存，显示This plugin does not support propagateSizeHints() This plugin does not support raise()想请教一下您有没有解决方法。同时请问只有训练文件 没有预测文件吗？

Hi, can you share about the data preprocessing, as an example of monuseg, how to get the labels in png form from the original xml file?