DynUNet configuration to reproduce BraTS2020 nnUnet baseline's architecture

[LouisDo2108]

Hi! I am using MONAI to reimplement the nnU-Net baseline's architecture from this paper: https://arxiv.org/pdf/2011.00848.pdf
Here is the overview of the architecture.

My current configuration is

kernel_size = [3]*6
strides = [1] + [2]*5
net = DynUNet(
    spatial_dims=3,
    in_channels=4,
    out_channels=4,
    kernel_size=kernel_size,
    strides=strides,
    upsample_kernel_size=[1]*5,
    norm_name="instance",
    deep_supervision=True,
    deep_supr_num=3,
)

Is this correct? Thank you!

[KumoLiu]

Hi @yiheng-wang-nv, could you please share some comments here? Thanks in advance!