Skip to content

Differentiate between model output and postprocessing output #6724

Closed

Description

@wyli

The metadata states what the network produces as its output but if the postprocessing transform sequence does a lot to change the shape and format of this output there will be a disconnect between what is expected and the actual results. One example is a bundle which produces a certain number of channels that don't directly relate to classes which is then transformed into an output with a different number of layers representing one-hot encoded multiclass labels

  • update spec
  • update existing bundles output spec (model zoo)
  • downstream app. (monai label) reading output spec for bundle validation

Activity

  1. converted this from a draft issue on Jul 14, 2023
  2. added this to MONAI v1.4 and removed this from MONAI v1.3on Oct 13, 2023
  3. moved this to Todo in MONAI v1.4on Oct 13, 2023
  4. moved this from Todo to No status in MONAI v1.4on Nov 3, 2023
  5. moved this to Todo in MONAI v1.4on Dec 1, 2023
  6. self-assigned this
    on Sep 10, 2024
  7. ericspod commented on Sep 10, 2024

    @ericspod
    Member

    We have updated the schema to support this, there are a few bundles which should be changed to reflect this difference, but also MONAILabel and other tools require some changes to correctly pick up which result they expect.

  8. added this to MONAI v1.5 and removed this from MONAI v1.4on Sep 20, 2024
  9. moved this to In progress in MONAI v1.5on Sep 20, 2024
  10. moved this from In progress to Done in MONAI v1.5on Jun 6, 2025
  11. closed this as completedby moving to Done in MONAI v1.5on Jun 6, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Type

No type

Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions