Skip to main content
. Author manuscript; available in PMC: 2022 Mar 3.
Published in final edited form as: IEEE Trans Med Imaging. 2022 Mar 2;41(3):543–558. doi: 10.1109/TMI.2021.3116879

Fig. 4.

Fig. 4.

U-Net architecture of ϕθ = hθ(m, f). Each block of the encoder features a 3D convolution with n = 256 filters and a LeakyReLU layer (0.2). Stride-2 convolutions each halve the resolution relative to the input. In the decoder, each convolution is followed by an upsampling layer and a skip connection (long arrows). The SVF vθ is obtained at half resolution, yielding the warp ϕθ after integration and upsampling. All kernels are of size 3 × 3 × 3. The final layer uses n= 3 filters.