timing analysis on the RnR-ExM challenge
Closed this issue · 3 comments
Hi bigstream team,
Thank you for releasing the code, and congrats on winning the first place in the RnR-ExM challenge.
Do you have the code for the results or timing analysis of the method for the dataset?
Hi Rohit,
I wish there was a team - but it's just me 😄
All notebooks used for the rnr-exm challenge as well as a pdf of my slides are now included in the repository here: https://github.com/GFleishman/bigstream/tree/master/notebooks/rnr-exm_notebooks
Note the slides contained tons of movies showing the alignment results which obviously aren't rendered in the pdf.
For timing I just used the Jupyter Notebook %%time cell magic. For time comparisons make sure you are running on a comparable machine. My times were on a machine with 16 2.7GHz Intel SkyLake Platinum 8168 cores running Oracle Linux 8.3. All times are entirely CPU based, in fact the machine I ran on doesn't have a GPU. It may also be important to determine the exact bigstream version number used for the rnr-exm challenge, since I've been working on it frequently.
Please consider clicking the star button for the bigstream repository. I'm working on it intensively and lots of new updates are coming regularly. Even if you are just using it for benchmarking it helps to know who and how the software is being used.
Please say hello to Paul and Jim for me.
Thanks,
Greg
Thanks so much Greg.
Does your method also run on GPU? I'm working on something GPU-based so it will be fairer if I can include GPU comparisons as well.
Will definitely say hello 😃. In fact, Paul and Jim introduced me to your work - very cool stuff!
Bigstream is totally GPU free for the time being. Most of the alignment methods call down to SimpleITK. I haven't explored what GPU options are available in SimpleITK or ITK in Python. Interpolation and smoothing would definitely be faster on a GPU. Autograd could also be a big asset for certain registration tasks. Let me know if there is anything else I can help with 😄