Skip to content

Upgrade Augmentations Pipeline in OTX + 16bit images support#5021

Merged
leoll2 merged 1930 commits intodevelopfrom
kp/kornia
Mar 19, 2026
Merged

Upgrade Augmentations Pipeline in OTX + 16bit images support#5021
leoll2 merged 1930 commits intodevelopfrom
kp/kornia

Conversation

@kprokofi
Copy link
Contributor

@kprokofi kprokofi commented Nov 14, 2025 •
edited
Loading

Summary

resolves #5020

  • Provide batch augmentations support per model per task
  • Change augmentation pipeline to hybrid mode (batch augmentation + torchvision.v2). Operate only tensors
  • Remove all custom unnecessary augmentations
  • Provide DataAugmentationFactory to support configurable augmentation both for dataset based and batch based. This solution should work with Geti templates.
  • Provide 16bit images support for training via intensity transforms
  • Provide benchmark results

How to test

Checklist

  • The PR title and description are clear and descriptive
  • I have manually tested the changes
  • All changes are covered by automated tests
  • All related issues are linked to this PR (if applicable)
  • Documentation has been updated (if applicable)

@kprokofi kprokofi added the ALGO Any changes in OTX Algo Tasks implementation label Nov 14, 2025
@kprokofi kprokofi added this to the Geti Tune MVP milestone Nov 14, 2025
@kprokofi
Copy link
Contributor Author

kprokofi commented Nov 14, 2025 •
edited
Loading

First validation results:

Batch based augmentations + tensor only operations can improve iteration time almost 2x. The larger batch the larger improvements

Method Task Model iter_time
Current classification efficientnet_b0 0.446
Proposed classification efficientnet_b0 0.225
Current detection YOLOX_X 0.369
Proposed detection YOLOX_X 0.180

@leoll2 leoll2 removed this from the Geti Tune MVP milestone Nov 27, 2025
@github-actions github-actions bot added DEPENDENCY Any changes in any dependencies (new dep or its version) should be produced via Change Request on PM TEST Any changes in tests BUILD labels Jan 15, 2026
jpggvilaca and others added 23 commits January 20, 2026 12:40
@jpggvilaca
Add current training job UI (#5248)
ff7f2da
@leoll2
Update demo models (#5249)
822fcdd
@itallix @leoll2
Extend model endpoint with scores (#5254)
aacce3a 
Co-authored-by: Leonardo Lai <leonardo.lai@intel.com>
@A-Artemis
Add top picks to Model Architectures endpoint (#5255)
81f6b44
@JortBergfeld
Add query param for dataset revision to filter models on dataset (#5185)
36ed337 
Signed-off-by: Jort Bergfeld <jort.bergfeld@intel.com>
Co-authored-by: Copilot <198982749+Copilot@users.noreply.github.com>
@AlbertvanHouten
Update run.sh to be compatible with bash 3.x (#5257)
8173d44
@camiloHimura
Inference device selector (#5252)
7b7f428
@jpggvilaca
Integrate model variants API (#5251)
6a98280
@A-Artemis @jpggvilaca
Add model size calculation (#5259)
ba7ee13 
Co-authored-by: Joao Vilaca <joao.vilaca@intel.com>
@A-Artemis
Generate dataset item thumbnail on request (#5261)
79055bc
@dwesolow
fix: Project related issues (#5264)
6ee1b61
@jpggvilaca
Component tests for models screen (#5262)
cc84d29
@dwesolow
refactor(project): New project flow (#5260)
1c0ad34
@oep-renovate
chore(deps): update dependency lodash-es to v4.17.23 [security] (deve…
f001295 
…lop) (#5276)

Signed-off-by: oep-renovate[bot] <212772560+oep-renovate[bot]@users.noreply.github.com>
Co-authored-by: oep-renovate[bot] <212772560+oep-renovate[bot]@users.noreply.github.com>
@camiloHimura
Annotator label assignment (#5275)
e89f60f
@warrkan @leoll2
Video upload design (#5224)
542a45c 
Co-authored-by: Leonardo Lai <leonardo.lai@intel.com>
@itallix
Progress reporting in the training job (#5292)
f34a7b9
@leoll2
Upgrade PyTorch to 2.10 (#5286)
9bbdc3f
@leoll2
Upgrade OpenVINO and Model API (#5295)
98ebed4
@dwesolow
feat: Display recommended architectures and simplify code (#5291)
26d905d
@jpggvilaca
API integration pt3 (#5290)
281dc30
@camiloHimura
Annotator remove labels (#5296)
44444c7
@warrkan @jpggvilaca
Media table, separation of media and dataset items (#5294)
3d64e9b 
Co-authored-by: Joao Vilaca <joao.vilaca@intel.com>
@kprokofi kprokofi added this to the Geti 3.0 milestone Mar 17, 2026
@github-actions github-actions bot added the Geti Tune Backend Issues related to Geti Tune backend label Mar 17, 2026
@github-actions
Copy link

github-actions bot commented Mar 17, 2026 •
edited
Loading

📊 Test coverage report

Metric Coverage
Lines 44.2%
Functions 77.5%
Branches 87.3%
Statements 44.2%

@github-actions
Copy link

github-actions bot commented Mar 17, 2026 •
edited
Loading

Docker Image Sizes

CPU

Image Size
geti-cpu:pr-5021 1.01G
geti-cpu:sha-6e19405 1.01G

CUDA

Image Size
geti-cuda:pr-5021 5.55G
geti-cuda:sha-6e19405 5.55G

XPU

Image Size
geti-xpu:pr-5021 3.34G
geti-xpu:sha-6e19405 3.34G

@AlbertvanHouten AlbertvanHouten requested a review from Copilot March 19, 2026 10:20
Copilot AI reviewed Mar 19, 2026
View reviewed changes
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Copilot reviewed 193 out of 217 changed files in this pull request and generated 5 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

You can also share your feedback on Copilot code review. Take the survey.

leoll2
leoll2 previously approved these changes Mar 19, 2026
View reviewed changes
@kprokofi kprokofi dismissed stale reviews from leoll2 and AlbertvanHouten via 39bf5a5 March 19, 2026 12:20
@kprokofi kprokofi added this pull request to the merge queue Mar 19, 2026
@github-merge-queue github-merge-queue bot removed this pull request from the merge queue due to failed status checks Mar 19, 2026
@leoll2 leoll2 merged commit fb292b0 into develop Mar 19, 2026
51 checks passed
@leoll2 leoll2 deleted the kp/kornia branch March 19, 2026 16:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

Copilot code review Copilot Copilot left review comments

@leoll2 leoll2 leoll2 approved these changes

@AlbertvanHouten AlbertvanHouten AlbertvanHouten approved these changes

Assignees

No one assigned

Labels

ALGO Any changes in OTX Algo Tasks implementation BUILD DEPENDENCY Any changes in any dependencies (new dep or its version) should be produced via Change Request on PM Geti Tune Backend Issues related to Geti Tune backend TEST Any changes in tests

Projects

None yet

Milestone

Geti 3.0

Development

Successfully merging this pull request may close these issues.

Support 16-bit images in OTX Optimize OTX Data Augmentations pipeline