Refactor OTX Models

[kprokofi]

Summary

• Combine Explainable models with base models
• divide classification tasks for better maintenance and cleaner code
• Remove get_classification_layers and necessity to build models 3 time every time. Move this to incremental checkpoint loading
• Introduce DataInputParams, incapsulate input_size, mean and scale. Derive mean and std values from Normilize transform and automatically pass it to the model
• Align all models with same interface, introduce model_name for every model. Note: didn't pay much attention to Anomaly models since we will move them to separate Engine. Same for OV models (we will remove them)

How to test

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have ran e2e tests and there is no issues.
• I have added the description of my changes into CHANGELOG in my target branch (e.g., CHANGELOG in develop).​
• I have updated the documentation in my target branch accordingly (e.g., documentation in develop).
• I have linked related issues.

License

• I submit my code changes under the same Apache License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2025 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[ashwinvaidya17]

Thanks for the large refactor! I have one minor comment and one concern related to the anomaly models.

[sovrasov]

Thanks for the PR!
Is MaskDINO model also deleted?
Can we separate cleaning PRs and refactoring PRs to avoid huge changesets which are hard to review?

[kprokofi]

Thanks for the PR! Is MaskDINO model also deleted? Can we separate cleaning PRs and refactoring PRs to avoid huge changesets which are hard to review?

I understand that there are a lot of changes. However, I decided to delete models, that we rejected and those which are not exposed to Geti. My changes with input size affected all the models and it is better to delete them at all rather than fixing them and tests.

What has been deleted: YOLOV9, MaskDINO, HuggingFace models (we will move HuggingFace functionality to the separate engine) + CLI install (There was no need to delete this file in this PR, but if it is OK, I would not return it back)

[sovrasov]

Thanks for the PR! Is MaskDINO model also deleted? Can we separate cleaning PRs and refactoring PRs to avoid huge changesets which are hard to review?

I understand that there are a lot of changes. However, I decided to delete models, that we rejected and those which are not exposed to Geti. My changes with input size affected all the models and it is better to delete them at all rather than fixing them and tests.

What has been deleted: YOLOV9, MaskDINO, HuggingFace models (we will move HuggingFace functionality to the separate engine) + CLI install (There was no need to delete this file in this PR, but if it is OK, I would not return it back)

Unused models could have been deleted in advance, before the actual refactoring, like Ashwin did with a number of other models, or even tasks

[kprokofi]

Thanks for the PR! Is MaskDINO model also deleted? Can we separate cleaning PRs and refactoring PRs to avoid huge changesets which are hard to review?

I understand that there are a lot of changes. However, I decided to delete models, that we rejected and those which are not exposed to Geti. My changes with input size affected all the models and it is better to delete them at all rather than fixing them and tests.
What has been deleted: YOLOV9, MaskDINO, HuggingFace models (we will move HuggingFace functionality to the separate engine) + CLI install (There was no need to delete this file in this PR, but if it is OK, I would not return it back)

Unused models could have been deleted in advance, before the actual refactoring, like Ashwin did with a number of other models, or even tasks

Agree, I will create a separate PR with deleting things and update this one after

[kprokofi]

Cleaning PR: #4247

[kprokofi]

PR is updated and ready for review

[codecov]

Codecov Report

Attention: Patch coverage is 77.43590% with 264 lines in your changes missing coverage. Please review.

Project coverage is 80.43%. Comparing base (31823cd) to head (9e02d07).
Report is 136 commits behind head on develop.

Files with missing lines	Patch %	Lines
src/otx/core/model/hlabel_classification.py	54.19%	60 Missing ⚠️
src/otx/algo/instance_segmentation/maskrcnn.py	20.37%	43 Missing ⚠️
...c/otx/algo/classification/multiclass_models/vit.py	80.74%	26 Missing ⚠️
src/otx/core/model/multiclass_classification.py	74.24%	17 Missing ⚠️
src/otx/core/model/multilabel_classification.py	75.00%	15 Missing ⚠️
src/otx/core/model/base.py	69.04%	13 Missing ⚠️
.../algo/classification/hlabel_models/mobilenet_v3.py	80.00%	12 Missing ⚠️
src/otx/algo/classification/hlabel_models/vit.py	80.35%	11 Missing ⚠️
...c/otx/algo/classification/multilabel_models/vit.py	80.39%	10 Missing ⚠️
...o/classification/multilabel_models/mobilenet_v3.py	88.88%	5 Missing ⚠️
... and 19 more

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #4241      +/-   ##
===========================================
+ Coverage    77.11%   80.43%   +3.32%     
===========================================
  Files          308      269      -39     
  Lines        30571    24999    -5572     
===========================================
- Hits         23576    20109    -3467     
+ Misses        6995     4890    -2105     

Flag	Coverage Δ
py310	79.80% <77.43%> (+2.69%)	⬆️
py311	80.41% <77.43%> (+3.36%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.