Skip to content

feat(postprocess): add MEBin post-processing method #3116

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.

Sign up for GitHub

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

Jump to bottom
Open
wants to merge 23 commits into
base: main
Choose a base branch
from
Open

feat(postprocess): add MEBin post-processing method #3116

wants to merge 23 commits into from
+509 −4

Conversation

@StarPlatinum7
Copy link

@StarPlatinum7 StarPlatinum7 commented Nov 14, 2025 •
edited
Loading

📝 Description

This pull request introduces support for the MEBIN post-processing method within the Anomalib framework. The implementation includes a new post-processing algorithm that ​​mitigates false positives and false negatives in the binarization results by identifying the threshold interval where connected components remains stable.​ This strategy is proposed in AnomalyNCD (CVPR2025).

✨ Changes

Select what type of change your PR is:

  • 🚀 New feature (non-breaking change which adds functionality)
  • 🐞 Bug fix (non-breaking change which fixes an issue)
  • 🔄 Refactor (non-breaking change which refactors the code base)
  • ⚡ Performance improvements
  • 🎨 Style changes (code style/formatting)
  • 🧪 Tests (adding/modifying tests)
  • 📚 Documentation update
  • 📦 Build system changes
  • 🚧 CI/CD configuration
  • 🔧 Chore (general maintenance)
  • 🔒 Security update
  • 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)

✅ Checklist

Before you submit your pull request, please make sure you have completed the following steps:

  • 📚 I have made the necessary updates to the documentation (if applicable).
  • 🧪 I have written tests that support my changes and prove that my fix is effective or my feature works (if applicable).
  • 🏷️ My PR title follows conventional commit format.

For more information about code review checklists, see the Code Review Checklist.

ashwinvaidya17 and others added 21 commits November 14, 2025 13:42
@ashwinvaidya17 @StarPlatinum7
🚧 ci(tests): Use uv commands (open-edge-platform#2999)
0f4df8e 
* Use uv commands

Signed-off-by: Ashwin Vaidya <[email protected]>

* Add source activate step

Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@rajeshgangireddy @StarPlatinum7
🚀 feat(model): Make coreset selection for patchcore faster (open-edge…
c0ad43f 
…-platform#2968)

* 🚀 feat(model): make improvements for patchcore on xpu

* 🔄 refactor(k_center_greedy): remove profiling code.

Signed-off-by: rajeshgangireddy <[email protected]>

* 🚀 feat(benchmark): add seed setting for reproducibility in run_benchmark function

* 🔄 refactor(k_center_greedy): enhance device handling and remove unused benchmark scripts

* 🔄 refactor(k_center_greedy): remove the unused case of multiple centers  to simplify

* 🐞 fix(comments): update variable names

---------

Signed-off-by: rajeshgangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@oep-renovate @StarPlatinum7
chore(deps): update github actions (open-edge-platform#2983)
7f8d0a1 
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>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @code-dev05
@rajeshgangireddy @StarPlatinum7
🚀 feat(data): BMAD dataset (open-edge-platform#3000)
edb9f40 
* 🚀 feat(dataset): add BMAD dataset to Anomalib (open-edge-platform#2900)

* Adding BMAD dataset to Anomalib

Signed-off-by: Devansh Agarwal <[email protected]>

* Fixed pre-commit hook errors

Signed-off-by: Devansh Agarwal <[email protected]>

---------

Signed-off-by: Devansh Agarwal <[email protected]>

* 🔧 refactor(data): Polish BMAD dataset (open-edge-platform#2981)

* 🧹 Polish dataset

🧪 Add tests
📋 Fix documentation
🛠️  Minor fixes

Signed-off-by: Ashwin Vaidya <[email protected]>

* Update src/anomalib/data/datamodules/image/bmad.py

Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>

* Fix mask_samples

Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: Devansh Agarwal <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Devansh Agarwal <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @PaulMcHard
@samet-akcay @monzelr @rajeshgangireddy @StarPlatinum7
🚀 feat(dataset): add 3D-ADAM dataset to Anomalib (open-edge-platform#…
899a3fe 
…3003)

* 🚀 feat(dataset): add 3D-ADAM dataset to Anomalib (open-edge-platform#2986)

* 🐞 fix(utils): revert `OptionalImport` class to handle missing optional dependencies (open-edge-platform#2956)

* fix(data): enable pin_memory for DataLoader instances across the codebase

This commit updates various DataLoader instances in the project to enable the  option, enhancing performance for data loading on GPU. Changes were made in the following files:

- : Updated train and test DataLoader configurations.
- : Modified datamodule DataLoader to include .
- : Added  to evaluation DataLoader.
- : Updated DataLoader for datasets to utilize .
- : Enabled  for reference dataset DataLoader.
- : Adjusted inference DataLoader to include .

These changes aim to optimize memory usage and improve data transfer speeds during model training and inference.

Signed-off-by: samet-akcay <[email protected]>

* refactor(models): streamline decoder retrieval in  function

This commit refactors the  function in  to utilize a dictionary mapping for decoder architectures, improving readability and maintainability. The previous conditional checks have been replaced with a more efficient approach, enhancing the overall structure of the code.

Signed-off-by: samet-akcay <[email protected]>

* fix(model): update logit_scale initialization to use torch.log for consistency

This commit modifies the initialization of the logit_scale parameter in the CLIP model to utilize torch.log instead of np.log. This change ensures consistency in tensor operations and improves compatibility with PyTorch's computation graph.

Signed-off-by: samet-akcay <[email protected]>

* fix(model): update anomaly map generation to use torch tensors for calculations

This commit modifies the anomaly map generation logic to utilize PyTorch tensors instead of NumPy arrays for various calculations. This change enhances compatibility with the PyTorch computation graph and improves performance by leveraging GPU acceleration. Key updates include the conversion of statistical calculations and tensor operations to use PyTorch functions, ensuring consistency in tensor handling throughout the code.

Signed-off-by: samet-akcay <[email protected]>

* refactor(model): enhance anomaly map generation with PyTorch for statistical calculations

This commit refactors the anomaly map generation logic to replace NumPy-based statistical calculations with PyTorch equivalents, specifically using the  distribution for computing tau. Additionally, it improves precision handling by allowing the use of float64 in high precision mode. The changes streamline the computation process and maintain compatibility with the PyTorch computation graph.

Signed-off-by: samet-akcay <[email protected]>

* chore(license): update license year

Signed-off-by: samet-akcay <[email protected]>

* fix(download): enhance URL validation and update download logic

This commit improves the URL validation in the download function to ensure only http and https schemes are allowed. Additionally, it adds comments to clarify the safety of using  under these conditions, enhancing code readability and security awareness.

Signed-off-by: samet-akcay <[email protected]>

* 🗑️ chore(imports): remove OptionalImport class for optional dependency handling

Signed-off-by: samet-akcay <[email protected]>

* 🐛 fix(imports): Replace OptionalImport with dummy classes for missing dependencies

This commit removes the OptionalImport class and introduces dummy classes for various loggers and video utilities when their respective dependencies are not installed. The changes enhance error messaging by providing clear instructions for installation, improving user experience when encountering missing packages.

Affected files:
- video.py
- comet.py
- mlflow.py
- tensorboard.py
- wandb.py
- ollama.py

Signed-off-by: samet-akcay <[email protected]>

---------

Signed-off-by: samet-akcay <[email protected]>
Signed-off-by: Paul McHard <[email protected]>

* Added 3D-ADAM Dataset

Signed-off-by: Paul McHard <[email protected]>

* Added tests and example config

Signed-off-by: Paul McHard <[email protected]>

* 🔨 fix(semgrep): ignore uv.lock in semgrep (open-edge-platform#2972)

ignore uv.lock in semgrep

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Paul McHard <[email protected]>

* 🐞 fix(utils): Run CS Flow model in eval() so the weights of the model itself will not be modified (open-edge-platform#2966)

fix in dryrun_find_featuremap_dims: run the model in eval() so the weights of the model itself will not be modified

Signed-off-by: Paul McHard <[email protected]>

* 🐛 fix(ovinferencer): Fix openvino cache directory  (open-edge-platform#2979)

🐛 fix(ovinferencer): update cache folder name to 'openvino_cache' and ensure string path for set_property

Signed-off-by: Paul McHard <[email protected]>

* changes per PR and pre-commit checks

Signed-off-by: Paul McHard <[email protected]>

* resolve circular import issue

Signed-off-by: Paul McHard <[email protected]>

* Resolve DownloadInfo URL issue.

Signed-off-by: Paul McHard <[email protected]>

* Resolve DownloadInfo URL

Signed-off-by: Paul McHard <[email protected]>

* restore doc images ignored in error

Signed-off-by: Paul McHard <[email protected]>

* Removed prettier cache included in error, re-run pre-commit hooks

Signed-off-by: Paul McHard <[email protected]>

* explictly resolve EOF error on docs/source/examples

Signed-off-by: Paul McHard <[email protected]>

* resolve filename issue in adam3d test.

Signed-off-by: Paul McHard <[email protected]>

* Restore uv.lock

Signed-off-by: Paul McHard <[email protected]>

---------

Signed-off-by: samet-akcay <[email protected]>
Signed-off-by: Paul McHard <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Paul McHard <[email protected]>
Co-authored-by: Samet Akcay <[email protected]>
Co-authored-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Rune <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>

* revert changes to example

Signed-off-by: Ashwin Vaidya <[email protected]>

* revert file permissions

Signed-off-by: Ashwin Vaidya <[email protected]>

* 🔧 refactor(data): Polish 3dAdam dataset  (open-edge-platform#3002)

* Add changelog + fix dataset location

Signed-off-by: Ashwin Vaidya <[email protected]>

* revert changes to example

Signed-off-by: Ashwin Vaidya <[email protected]>

* revert file permissions

Signed-off-by: Ashwin Vaidya <[email protected]>

* Fix tests

Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: Ashwin Vaidya <[email protected]>

* revert examples

Signed-off-by: Ashwin Vaidya <[email protected]>

* revert file permissions

Signed-off-by: Ashwin Vaidya <[email protected]>

* Add imagecodecs dependency

Signed-off-by: Ashwin Vaidya <[email protected]>

* Apply suggestion from @samet-akcay

Co-authored-by: Samet Akcay <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: samet-akcay <[email protected]>
Signed-off-by: Paul McHard <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Paul McHard <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Paul McHard <[email protected]>
Co-authored-by: Samet Akcay <[email protected]>
Co-authored-by: Rune <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@samet-akcay @ashwinvaidya17
@StarPlatinum7
🔧 chore(release): prepare v2.2.0 release (open-edge-platform#2960)
4e86d37 
* fix(data): enable pin_memory for DataLoader instances across the codebase

This commit updates various DataLoader instances in the project to enable the  option, enhancing performance for data loading on GPU. Changes were made in the following files:

- : Updated train and test DataLoader configurations.
- : Modified datamodule DataLoader to include .
- : Added  to evaluation DataLoader.
- : Updated DataLoader for datasets to utilize .
- : Enabled  for reference dataset DataLoader.
- : Adjusted inference DataLoader to include .

These changes aim to optimize memory usage and improve data transfer speeds during model training and inference.

Signed-off-by: samet-akcay <[email protected]>

* refactor(models): streamline decoder retrieval in  function

This commit refactors the  function in  to utilize a dictionary mapping for decoder architectures, improving readability and maintainability. The previous conditional checks have been replaced with a more efficient approach, enhancing the overall structure of the code.

Signed-off-by: samet-akcay <[email protected]>

* fix(model): update logit_scale initialization to use torch.log for consistency

This commit modifies the initialization of the logit_scale parameter in the CLIP model to utilize torch.log instead of np.log. This change ensures consistency in tensor operations and improves compatibility with PyTorch's computation graph.

Signed-off-by: samet-akcay <[email protected]>

* fix(model): update anomaly map generation to use torch tensors for calculations

This commit modifies the anomaly map generation logic to utilize PyTorch tensors instead of NumPy arrays for various calculations. This change enhances compatibility with the PyTorch computation graph and improves performance by leveraging GPU acceleration. Key updates include the conversion of statistical calculations and tensor operations to use PyTorch functions, ensuring consistency in tensor handling throughout the code.

Signed-off-by: samet-akcay <[email protected]>

* refactor(model): enhance anomaly map generation with PyTorch for statistical calculations

This commit refactors the anomaly map generation logic to replace NumPy-based statistical calculations with PyTorch equivalents, specifically using the  distribution for computing tau. Additionally, it improves precision handling by allowing the use of float64 in high precision mode. The changes streamline the computation process and maintain compatibility with the PyTorch computation graph.

Signed-off-by: samet-akcay <[email protected]>

* chore(license): update license year

Signed-off-by: samet-akcay <[email protected]>

* fix(download): enhance URL validation and update download logic

This commit improves the URL validation in the download function to ensure only http and https schemes are allowed. Additionally, it adds comments to clarify the safety of using  under these conditions, enhancing code readability and security awareness.

Signed-off-by: samet-akcay <[email protected]>

* 📝 chore(changelog): update changelog for v2.2.0 release with added features and fixes

* 🐛 fix(version): update version from 2.2.0dev to 2.2.0

* Update CHANGELOG.md

Signed-off-by: Ashwin Vaidya <[email protected]>

* Update changelog

Signed-off-by: Ashwin Vaidya <[email protected]>

* Use full PR links

Signed-off-by: Ashwin Vaidya <[email protected]>

* Add authors in changelog

Signed-off-by: Ashwin Vaidya <[email protected]>

* Update CHANGELOG.md

Co-authored-by: Copilot <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: samet-akcay <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Copilot <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @StarPlatinum7
🔧 docs(changelog): Fix entries (open-edge-platform#3008)
edfb369 
* Remove duplicate entry

Signed-off-by: Ashwin Vaidya <[email protected]>

* Add missing entry

Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @StarPlatinum7
🔧 chore(version): Update main version to dev (open-edge-platform#3010)
c265c7a 
Update version

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@rajeshgangireddy @StarPlatinum7
📦 docs(README): Update ReadMe for V2.2.0 release (open-edge-platform#…
e4ae43b 
…3017)

📦 update(README): Announce v2.2.0 release with new datasets, metrics, and performance improvements

Signed-off-by: StarPlatinum7 <[email protected]>
@maxxgx @StarPlatinum7
🐛 fix(metrics): disable mps for torch metrics (open-edge-platform#3019)
595a008 
🐛 fix(metrics): disable mps for torch metrics (open-edge-platform#3018)


* move to cpu when device is mps



---------

Signed-off-by: Ma, Xiangxiang <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @StarPlatinum7
🔧 chore(codeowners): Update owners (open-edge-platform#3042)
89eb295 
Update CODEOWNERS

Signed-off-by: Ashwin Vaidya <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@mramotowski @ashwinvaidya17
@StarPlatinum7
chore: add code owners for GitHub workflows and security policy (open…
fdb6e7c 
…-edge-platform#3046)

* chore: add code owners for GitHub workflows and security policy

- Add @kamillipka @mramotowski @ivanzati @AlexanderBarabanov as owners for .github/workflows/
- Add same team as owners for SECURITY.md

Signed-off-by: mramotowski <[email protected]>

* Apply suggestion from @Copilot

Co-authored-by: Copilot <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>

---------

Signed-off-by: mramotowski <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Copilot <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@alexriedel1 @rajeshgangireddy @StarPlatinum7
🚀 feat(model): Enable Patchcore Training Half Precision (open-edge-pl…
4708448 
…atform#3031)

patchcore infer dtype from model dtype

Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@blaz-r @rajeshgangireddy @StarPlatinum7
🚀 feat(model): Updated SuperSimpleNet to latest version (open-edge-pl…
46481ec 
…atform#3036)

* Fix squeeze on 1dim score

Signed-off-by: blaz.rolih <[email protected]>

* Add option to train without masks

Signed-off-by: blaz.rolih <[email protected]>

* Add JIMS separate feat extension

Signed-off-by: blaz.rolih <[email protected]>

* Update docs for JIMS extension

Signed-off-by: blaz.rolih <[email protected]>

* Remove unused get_params method

Signed-off-by: blaz.rolih <[email protected]>

* Add unit tests for SSN

Signed-off-by: blaz.rolih <[email protected]>

* Rename vars and update metrics in readme

Signed-off-by: blaz.rolih <[email protected]>

---------

Signed-off-by: blaz.rolih <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@mramotowski @StarPlatinum7
ci: migrate PyPI publishing to OIDC authentication (open-edge-platfor…
9e5610a 
…m#3054)

cv1.0.0i: migrate PyPI publishing to OIDC authentication

Replace token-based authentication with OpenID Connect (OIDC) for
enhanced security when publishing to PyPI and Test PyPI.

Signed-off-by: mramotowski <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@ashwinvaidya17 @alexriedel1
@rajeshgangireddy @StarPlatinum7
🚀 feat(model): Enable Patchcore Training Half Precision (open-edge-pl…
4798892 
…atform#3047) (open-edge-platform#3055)

* 🚀 feat(model): REFACTOR Enable Patchcore Training Half Precision (open-edge-platform#3047)

* patchcore infer dtype from model dtype

* lock

* pre-commits

Signed-off-by: Alexander Riedel <[email protected]>

---------

Signed-off-by: Alexander Riedel <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>

---------

Signed-off-by: Alexander Riedel <[email protected]>
Signed-off-by: Ashwin Vaidya <[email protected]>
Co-authored-by: Alexander Riedel <[email protected]>
Co-authored-by: Rajesh Gangireddy <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@abc-125 @StarPlatinum7
🐞 fix(data): Fixed comparison of file extensions for Folder dataset (o…
903d24f 
…pen-edge-platform#3059)

* Fixed Folder dataset to use upper case file extension

* Convert provided extensions to lower case

Signed-off-by: StarPlatinum7 <[email protected]>
@oep-renovate @AlexanderBarabanov @StarPlatinum7
chore(deps): lock file maintenance (open-edge-platform#3014)
53d46ac 
* chore(deps): lock file maintenance

Signed-off-by: oep-renovate[bot] <212772560+oep-renovate[bot]@users.noreply.github.com>

* use semgrep action from geti-ci

Signed-off-by: Barabanov <[email protected]>

* fix pre-commit

Signed-off-by: Barabanov <[email protected]>

* update sha

Signed-off-by: Barabanov <[email protected]>

* remove local semgrep action

Signed-off-by: Barabanov <[email protected]>

* update sha

Signed-off-by: Barabanov <[email protected]>

* update semgrepignore

Signed-off-by: Barabanov <[email protected]>

---------

Signed-off-by: oep-renovate[bot] <212772560+oep-renovate[bot]@users.noreply.github.com>
Signed-off-by: Barabanov <[email protected]>
Co-authored-by: oep-renovate[bot] <212772560+oep-renovate[bot]@users.noreply.github.com>
Co-authored-by: Barabanov <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@AlexanderBarabanov @StarPlatinum7
fix(ci): disable torch upgrade in Renovate config (open-edge-platfo…
98160c9 
…rm#3088)

update renovate config

Signed-off-by: Barabanov <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@waschsalz @StarPlatinum7
fix(accelerator): Adding name method in XPUAccelerator (open-edge-pla…
c4d9a95 
…tform#3108)

* Update xpu.py regarind PR open-edge-platform#3092

Added the name method to fix an issue related to a newly added feature in lightning 2.5.6

Signed-off-by: Niclas <[email protected]>

* Update xpu.py

Signed-off-by: Niclas <[email protected]>

* Update xpu.py

Signed-off-by: Niclas <[email protected]>

* Update xpu.py with docstring

Signed-off-by: Niclas <[email protected]>

* Update xpu.py with correct docstring

Signed-off-by: Niclas <[email protected]>

* added name method for XPUAccelerator

Signed-off-by: waschsalz <[email protected]>

---------

Signed-off-by: Niclas <[email protected]>
Signed-off-by: waschsalz <[email protected]>
Signed-off-by: StarPlatinum7 <[email protected]>
@StarPlatinum7
mebin
3267d33 
Signed-off-by: StarPlatinum7 <[email protected]>
@StarPlatinum7 StarPlatinum7 changed the title feat(postprocess): add MEBin post-processing method #2928 feat(postprocess): add MEBin post-processing method Nov 14, 2025
@StarPlatinum7 StarPlatinum7 changed the title feat(postprocess): add MEBin post-processing method feat(postprocess): add MEBin post-processing method Nov 14, 2025
samet-akcay
samet-akcay requested changes Nov 24, 2025
View reviewed changes
Copy link
Contributor

@samet-akcay samet-akcay left a comment

Choose a reason for hiding this comment

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

Thanks for this great contribution! Overall, it looks great. I have few comments here and there.

@rajeshgangireddy, would it be an idea to merge this to a feature branch, and address the issues ourselves?

src/anomalib/metrics/threshold/mebin.py
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

there are some formatting issues in this file, which could be addressed by running pre-commit run -a

Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

Also, this file needs annotations in the class/functions etc?

src/anomalib/metrics/threshold/mebin.py
"""

import cv2
import numpy as np
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

would it be possible to use torch instead of numpy? We would want everything to be a torch operation end-to-end

src/anomalib/metrics/threshold/mebin.py
>>> binarized_maps, thresholds = mebin.binarize_anomaly_maps()
"""

def __init__(self, anomaly_map_path_list=None, sample_rate=4, min_interval_len=4, erode=True):
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

What if we want to use this post-processing in an end-to-end operation in anomalib? Would we need to follow this logic?

  1. Save anomaly maps to file
  2. Load them here and apply post-processing?

Would it not be possible to use this approach without loading the heatmaps from file?

src/anomalib/metrics/threshold/mebin.py

self.anomaly_map_path_list = anomaly_map_path_list
# Load anomaly maps as grayscale images if paths are provided
self.anomaly_map_list = [cv2.imread(x, cv2.IMREAD_GRAYSCALE) for x in self.anomaly_map_path_list]
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

if we have to load from filesystem, maybe we could use anomalib's built-in read_image util function?

src/anomalib/metrics/threshold/mebin.py

def get_threshold(self, anomaly_num_sequence, min_interval_len):
"""
Find the 'stable interval' in the anomaly region number sequence.
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

Suggested change
Find the 'stable interval' in the anomaly region number sequence.
Find the `stable interval` in the anomaly region number sequence.

src/anomalib/metrics/threshold/mebin.py
"""
interval_result = {}
current_index = 0
while current_index < len(anomaly_num_sequence):
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

are we certain that nested while loops are the most efficient way of running this approach?

src/anomalib/post_processing/mebin_post_processor.py
"""
anomaly_maps = predictions.anomaly_map
if isinstance(anomaly_maps, torch.Tensor):
anomaly_maps = anomaly_maps.detach().cpu().numpy()
Copy link
Contributor

@samet-akcay samet-akcay Nov 24, 2025

Choose a reason for hiding this comment

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

I see that we keep switching between numpy and torch. Would it be possible to just use torch instead?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ashwinvaidya17 ashwinvaidya17 Awaiting requested review from ashwinvaidya17 ashwinvaidya17 is a code owner

@rajeshgangireddy rajeshgangireddy Awaiting requested review from rajeshgangireddy rajeshgangireddy is a code owner

@mramotowski mramotowski Awaiting requested review from mramotowski

@AlexanderBarabanov AlexanderBarabanov Awaiting requested review from AlexanderBarabanov

@samet-akcay samet-akcay Awaiting requested review from samet-akcay samet-akcay is a code owner

Requested changes must be addressed to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.