search

microsoft / Recursive-Cascaded-Networks

[ICCV 2019] Recursive Cascaded Networks for Unsupervised Medical Image Registration
https://arxiv.org/abs/1907.12353
MIT License
362 stars 87 forks source link

about VTN network structure #18

Closed zzx332 closed 4 years ago

zzx332 commented 4 years ago

hi,I have some questions about the code of VTN

        pred6 = convolve('pred6', conv6_1, dims, 3, 1)
        upsamp6to5 = upconvolve('upsamp6to5', pred6, dims, 4, 2, shape5[1:4])
        deconv5 = upconvolveLeakyReLU(
            'deconv5', conv6_1, shape5[4], 4, 2, shape5[1:4])
        concat5 = tf.concat([conv5_1, deconv5, upsamp6to5], 4, 'concat5')

in the VTN code,every upsampling part have pred,this structure is not mentioned in the paper. what is the use of this operation?

zsyzzsoft commented 4 years ago

It would facilitate deep supervision but we didn't do it so I don't think it makes much difference. VTN should work without them.