search

adler-j / learned_primal_dual

Learned Primal-Dual Reconstruction
https://arxiv.org/abs/1707.06474
98 stars 37 forks source link

Reference TV reconstruction for ellipse data #3

Closed ChengV0 closed 4 years ago

ChengV0 commented 5 years ago

when i run reference_tv_shepp.py, image image image image Thanks for your reply.

AceCoooool commented 5 years ago

You can change the position of g and f in odl.solvers.pdhg. (However, the results is worst than paper)

ChengV0 commented 5 years ago

You can change the position of g and f in odl.solvers.pdhg. (However, the results is worst than paper)

thanks for your reply ,i will try it later.

adler-j commented 5 years ago

This is due to a standardization of parameters in ODL, I have a bugfix coming. The correct fix should be to rename g <-> f, if you just change the order then the steps etc are wrong. I though I had uploaded this but it seems I didn't.

ChengV0 commented 5 years ago

This is due to a standardization of parameters in ODL, I have a bugfix coming. The correct fix should be to rename g <-> f, if you just change the order then the steps etc are wrong. I though I had uploaded this but it seems I didn't.

thanks for your ad

You can change the position of g and f in odl.solvers.pdhg. (However, the results is worst than paper)

thanks for your reply ,i will try it later.

image this is the result after change the position of g and f.

ChengV0 commented 5 years ago

This is due to a standardization of parameters in ODL, I have a bugfix coming. The correct fix should be to rename g <-> f, if you just change the order then the steps etc are wrong. I though I had uploaded this but it seems I didn't. thanks for your advice,i will try it later.

adler-j commented 5 years ago

23,5 sounds very wrong. I'll have to look into that.

ChengV0 commented 5 years ago

23,5 sounds very wrong. I'll have to look into that.

Okay,waiting for your results

ChengV0 commented 5 years ago

23,5 sounds very wrong. I'll have to look into that.

Dear Adler, how about the result?

99991 commented 1 year ago

I am also interested. I think this issue should be reopened.