Extend image augmentation module and definitely fix device bug#32
Merged
PeterCalifano merged 25 commits intomainfrom Jan 3, 2026
Merged
Extend image augmentation module and definitely fix device bug#32PeterCalifano merged 25 commits intomainfrom
PeterCalifano merged 25 commits intomainfrom
Conversation
- Add new config entry to deactivate random apply (default selection) - Requires modification of random apply due to random order (previously misunderstood behaviour)
- Move method to static and review - Start some refactoring
- Minor improvements in augs helper - Start implementation of BorderAwareRandomAffine to replace RandomAffine
- Implement test for border crossing detection method - Rework class methods to comply with kornia Geometric Augmentation 2D base class - Minor fixes of output shapes
- Change interface to add new input options - Add assertions on parameters validity - Fix incorrect indexing (not needed for intensity augs)
- Reimplement factory function for AugsSequential from legacy function - Add interfaces in augs helper class for SoftBinarize module - Complete implementation of RandomSoftBinarizeImage augmentation module (blended binarization) with major bug fixes - Add implementation of unit test and general improvements
- Improves robustness to spurious 1s in the background (as caused by Blender)
- Add assertion on execution device for module - Add timing test to compare cpu vs gpu execution
- Add methods to resolve and move to device - Add move to device wherever relevant - Fix implementation of "to" method of the helper class - Implement tests and timing benchmark in test module
Contributor
There was a problem hiding this comment.
Pull request overview
This PR extends the image augmentation module with new capabilities, fixes device management issues, and adds comprehensive tests. The changes enable border-aware affine transformations for space images, introduce soft-binarization for texture removal, and ensure correct device placement throughout the augmentation pipeline.
Key Changes:
- Implemented
BorderAwareRandomAffinewith border crossing detection to prevent unrealistic transformations - Added
RandomSoftBinarizeImageaugmentation for texture content reduction - Fixed device management bugs in
ImageAugmentationsHelperto properly handle CPU/CUDA placement - Restructured code by moving augmentation classes to dedicated modules under
noise_models/
Reviewed changes
Copilot reviewed 9 out of 11 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| tests/datasets/test_ImagesAugmentation.py | Added comprehensive tests for new augmentations, device handling, and CPU vs CUDA timing benchmarks |
| pyTorchAutoForge/datasets/noise_models/OpticsErrorsAugs.py | New module containing optics-related augmentation imports and base setup |
| pyTorchAutoForge/datasets/noise_models/NoiseErrorsAugs.py | Relocated noise augmentation classes (PoissonShotNoise, RandomGaussianNoiseVariableSigma) |
| pyTorchAutoForge/datasets/noise_models/IntensityAugs.py | New intensity augmentations including RandomSoftBinarizeImage |
| pyTorchAutoForge/datasets/noise_models/GeometricAugs.py | New BorderAwareRandomAffine implementation with border detection logic |
| pyTorchAutoForge/datasets/auxiliary_functions.py | Updated pixel zero threshold from 7 to 1.0 |
| pyTorchAutoForge/datasets/ImagesAugmentation.py | Major refactoring: moved augmentation classes to separate modules, fixed device handling with new helper methods |
| lib/CommManager4MATLAB | Updated submodule commit reference |
| TODO | Updated task tracking for v0.2.X and v0.3.X milestones |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| # Overload "to" method | ||
| def to(self, *args, **kwargs): | ||
| """Overload to method to apply to all submodules.""" | ||
| device, _, _, _ = torch._C._nn._parse_to(*args, **kwargs) |
There was a problem hiding this comment.
Using private/internal PyTorch API torch._C._nn._parse_to. This may break in future PyTorch versions. Consider using public APIs or reimplementing device parsing logic.
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Owner
Author
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
List of changes for this PR: