Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Update wholeBody_ct_segmentation #397

Draft
wants to merge 4 commits into
base: dev
Choose a base branch
from

Conversation

KumoLiu
Copy link
Collaborator

@KumoLiu KumoLiu commented Apr 27, 2023

Description

While testing the lazy-resampling benchmark, I find that maybe there's room for improvement in the pre-processing.
In this PR, I mainly update several places below:

  • Remove NormalizeIntensityd since zeros aren't the background in the CT
  • Move ScaleIntensityd in front of the CropForeground and change the value to 0-1 since we can only pad zero now in the lazy mode.
  • Remove one-hot in the AsDiscreted which may save memory
  • Update the metric in the training part from Accuracy to Dice
  • Using the original label in the validation due to this issue

Note: the config was only tested on low resolution

cc @tangy5 @wyli

Status

Work in progress

Please ensure all the checkboxes:

  • Codeformat tests passed locally by running ./runtests.sh --codeformat.
  • In-line docstrings updated.
  • Update version and changelog in metadata.json if changing an existing bundle.
  • Please ensure the naming rules in config files meet our requirements (please refer to: CONTRIBUTING.md).
  • Ensure versions of packages such as monai, pytorch and numpy are correct in metadata.json.
  • Descriptions should be consistent with the content, such as eval_metrics of the provided weights and TorchScript modules.
  • Files larger than 25MB are excluded and replaced by providing download links in large_file.yml.
  • Avoid using path that contains personal information within config files (such as use /home/your_name/ for "bundle_root").

KumoLiu and others added 4 commits April 27, 2023 17:59
Signed-off-by: KumoLiu <yunl@nvidia.com>
Signed-off-by: KumoLiu <yunl@nvidia.com>
Signed-off-by: KumoLiu <yunl@nvidia.com>
@tangy5
Copy link
Collaborator

tangy5 commented Apr 28, 2023

Thanks for the updates, looking forward to seeing the complete updates.

@Nic-Ma
Copy link
Collaborator

Nic-Ma commented Jul 10, 2023

Hi @KumoLiu ,

Any update here?

Thanks.

@KumoLiu
Copy link
Collaborator Author

KumoLiu commented Jul 10, 2023

Hi @Nic-Ma, I don't have any update for this PR so far. The initial goal of this PR was to document what I thought could be improved in this model when I measured lazy resampling. However, because the data contained within this model zoo is actually high resolution, I was testing it on low resolution. So I'm just putting this PR here as a reference. If you think we need to keep updating it, I may need to redo some high-resolution experiments.
Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants