Jiancheng (JC) Yang

“We experimented with a shared DropBlock mask across different feature channels or each feature channel has its DropBlock mask. Algorithm 1 corresponds to the latter, which tends to work better...

我没有运行lfz的代码。如果你是完整且正确的运行他的代码，我觉得应该不会有问题，毕竟他的代码风格很清楚。如果你是自己处理的，那我认为是你可能没有做padding。他是取128的patch，所以需要在原始voxel上padding一圈64的。个人意见。

Got it, thx. To whom it may also concern, it should be [this line](https://github.com/lfz/DSB2017/blob/master/layers.py#L185)

Thank you! I will check this. Plus, it looks that the CT image sizes in the v1 and v2 are also different. Is there some special changes?

I guess you can refer to our example code about how to visualize the dataset, which provides a successful implementation of `montage` with pillow. You can also check how PIL...

Hi, Thanks for your response! I understand your concern on public data, but I still suggest to have some exceptions, for instance, OK to use torchvision / TF pretrained models,...

this issue appear to be alive after one year... [my implementation](https://gist.github.com/duducheng/e1a74ac8a1c45ac641b76ce0320cdb52) (based on this repo) is available for reference (together with Temporal Kernel RNN). **Caution** this is also one-year-ago code,...

Hi @RRouhi , Thanks for your interest ;) What do you mean by "custom data"? Sure you can always use your own data in PyTorch, right? Or do you want...