Björn Lütjens (he/him)
Björn Lütjens (he/him)
No. My assumption is that the method would need to be retrained on my dataset. I ended up matching the imagery via standard methods in opencv.
I also ran into issues with versions. The following changes made it run on my machine: To adapt tf2 to tf1: Change the import tf in VGG16.py and Registration.py to...
My machine has: ``` numpy==1.18.5 scipy==1.4.1 opencv-python-headless==4.4.0.42 tensorflow==2.3.0 tensorflow-estimator==2.3.0 lap==0.4.0 ```
okay, thanks! It seems like the PR you're referencing is actively being worked on, thanks!
+1; this seems like a really nice PR and I was also looking for this functionality.
+1 . torchvision.transforms.GaussianBlur seems significantly slower to me than the opencv equivalent.
Have you tried registering for the challenge at : https://competitions.codalab.org/competitions/29410#learn_the_details My guess is the data might be accessible for participants in the challenge. I just sent them a registration request,...
The proposed solution does not seem to work for me. The solution shows a progress bar for .load(), but not for open_mfdataset(), I think. I am running the code below....