search

dek924 / PerX2CT

Perspective Projection-Based 3D CT Reconstruction from Biplanar X-rays (ICASSP 2023, Best Student Paper Awards)
51 stars 10 forks source link

A bit question about feature extraction #2

Closed HB-pencil-zero closed 1 year ago

HB-pencil-zero commented 1 year ago

In ur paper , I saw " The local feature map Fvl and the global vector fvg are extracted fromthe third and fourth layer blocks, respectively."But in ur code , it seems the encoder will not change feature layer? I'm a bit confused about this question. Would u tell me why ? Thanks for ur goodwork . image

dek924 commented 1 year ago

It seems that the code you're referring to comes from the 'x2ct_nerf.modules.inr.model.PerspectiveINRNet' module. In this module, only local features are utilized, as specified by the 'PerX2CT' setting, which excludes global features. If you're interested in using both global and local features, please refer to the 'x2ct_nerf.modules.inr.model_global.PerspectiveConcatAfterINRGLNet' code.