Open heyblackC opened 2 years ago
Hi! Thanks for these kind words!
I have pushed the SbR models for the Allen_labels and BigBrain datasets. Feel free to use them and let me know if you find any issues running/loading, etc...
Regarding the number of images, the training of these networks is actually pixelwise, meaning that one could think of each pixel as an observation, but that are efficiently trained by learning the full image at each iteration. Moreover, we use geometric data augmentation to account for unseen deformations. I think that these numbers are ok for the application at hand.
Hello! Thanks for your reply. I wiil let you know if I encounter other issues.
For now, that's enough. Thanks.
Hello! I learn a lot from your paper, and I think you did a great job!
And I wonder if you can provide a best checkpoint of your SbR model for us? In case that we want to compare your model to ours. Maybe you can provide me a link to download this checkpoint? Both Allenlabel and BigBrain. Thanks!!
At last, I want to ask you if allenlabel dataset which just contain 93 pieces of images is too small for network to train? Do reviewers point this problem out?
Thank you!