Borda / BIRL

BIRL: Benchmark on Image Registration methods with Landmark validations

Home Page:http://borda.github.io/BIRL

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

method: DROP

Borda opened this issue · comments

transfer registration methods once used in previous registration benchmarks

Asked question to the authors about good parameter setting, because this takes hours for ANHIR dataset...

On Wed, 15 May 2019 at 10:28, Glocker, Ben ...@imperial.ac.uk wrote:

Dear Jiri,
I believe with your high-res images you are running into memory problems. The graph optimization that is used within drop is memory demanding.
One solution could be to tile your image into 1k by 1k images (or less) and then merge the resulting displacement fields back together.

Dear Ben,
may I have a kindly one more question, about some good parameter setting? I want to run your method on histology dataset with images of size about 10x10k pixels, but it seems that each registration takes hours.
Please see my parameters in the attached file. Also, the usage of your methods can be found here - https://github.com/Borda/BIRL/blob/DROP/bm_experiments/bm_DROP.py
Thank you very much for your advice

New DROP released and related discussion - biomedia-mira/drop2#2