This repository was archived by the owner on Apr 17, 2023. It is now read-only.
[Semi-SL/Detection] Refactoring mpa/det/ to support multiple learning technique#94
Merged
jaegukhyun merged 15 commits intootxfrom Dec 7, 2022
Merged
[Semi-SL/Detection] Refactoring mpa/det/ to support multiple learning technique#94jaegukhyun merged 15 commits intootxfrom
jaegukhyun merged 15 commits intootxfrom
Conversation
jaegukhyun
requested review from
JihwanEom,
kprokofi,
sovrasov and
supersoob
supersoob
reviewed
Dec 5, 2022
Contributor
Author
|
@supersoob @sovrasov @kprokofi @JihwanEom It's time to merge this PR, please review this PR. I want you to add your implementation for semi-sl from this branch. After merging this PR, I will open OTX PR, because semi-sl cli test requires this mpa branch |
supersoob
reviewed
Dec 5, 2022
supersoob
approved these changes
Dec 5, 2022
supersoob
left a comment
supersoob
left a comment
There was a problem hiding this comment.
Overall, it looks good to me. Let's wait for other's reviews. After merging this PR, I think it's gonna be good to make each PR for each task, referring this PR's code change. Thanks for your hard work, Jaeguk:)
goodsong81
reviewed
Dec 5, 2022
goodsong81
left a comment
goodsong81
left a comment
There was a problem hiding this comment.
I support @supersoob's point.
Why don't you create classincr directory?
Being a default option has nothing to do with class/directory structure.
Anyway, it's just a suggestion.
JihwanEom
reviewed
Dec 5, 2022
…y task adapt hook for semisl
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR includes refactoring mpa/det codes.
Current mpa/det's codes highly focus on class incremental learning, and this make hard to enable other training method.
To solve this issue, I split configure function in mpa/det/stage.py into two things. First is configuring configs for general fine tuning. Second is configuring configs for class-incremental learning.
And I created ClassIncrStage, which inherit base Stage, to support class incremental learning. Now trainer at mpa/det inherits from ClassIncrStage, because class incremental learning is our default recipe.
To enable semi-sl for detection, SemiStage has to be different with ClassIncrStage. Therefore I created SemiStage. SemiStage inherits from ClassIncrStage, and it modifies some configuration function for supporting semi-sl. Since it inherits ClassIncrStage, it can support class incremental scenarios.
As SemiStage is newly generated, Trainer and Inferencer for semi-sl is also generated, and they inherits SemiStage.
You can test this with otx branch for semi-sl(https://github.com/openvinotoolkit/training_extensions/tree/feature/otx-semisl-det). From this PR link(open-edge-platform/training_extensions#1398), you can find a description for this otx branch