8000 GitHub - jueqiw/pytorch_Unet
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

jueqiw/pytorch_Unet

Repository files navigation

regulation part

  • fix kernal size = 3 problem
  • writing fine-tune part
  • read the pytorch lightning pruning part
  • change the code to before
  • try the resample, padding and interplot method
  • try resample + cropping
  • do cropping and patch-based
  • adding data augmentation fine tune
  • adding one more layer without concat
  • try dice and BCE with weight
  • try denseUnet
  • Be more careful with the learning rate
  • try leave one out
  • select value module type = 1

later to do

  • write the predict part, and using the mp4 part
  • fix mp4 problem
  • update the mp4 quality
  • when test, the label do not do the data augmentation resize part
  • tune the data agumentation's parameters
  • fix dice score = inf problem
  • make the epoch more smaller / make the every_epoch_avl = fraction
  • write test part on the 1069 data
  • why there is a warning in mp4 part
  • make it into a pip package
  • look at the data after cropping and resize, and directly resize's difference?
  • why every time train, the train loss would suddenly go up a lot after a long time?

Reading

  • reading nnUnet again
  • read about nnUnet learning rate

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
0