Daniil Kazantsev
Daniil Kazantsev
Much more complicated phantoms and projection data can be generated by enabling cutoffs to the objects, e.g. half of the ellipse, gaussian etc.
* First complete #65 for at least for 2D case * Add the demo case of adding various levels of random noise to objects then using deep CNN to denoise...
Long term plans - to ensure ability to model dynamic behaviour of phantoms by providing descriptors (vectors for describing the motion). This insures a capability of creating complex motion patterns....
to implement: https://pubs.rsna.org/doi/suppl/10.1148/radiol.11101782/suppl_file/11101782appendix.pdf
Investigation and implementation of: * Baranov minimal projections criteria * Gerchberg-type initialisation (few-angles recovery) * Miltigrid reconstruction (recovery on lower scale, upsampling) * ?
Thanks a lot for refactoring guys! While I was fiddling with the toolkit to make it work for ToMoBAR, I noticed that the output volumes in `demo_gpu_regularisers3D.py` are not good...
CuPy requires for CUDA kernels to be present during the build. I've modified the `build.sh` and `bld.bat` files to do that. @paskino A question about a CuPy dependency. It can...