Heart-library

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Fixes alignment issue with augmentation (attack) and image classifier output. Both now return Sequence\[ArrayLike\] as per MAITE protocol.
- Fixes alignment of JaticPyTorchObjectDetector call method. Input type constrained to Sequence\[ArrayLike\] as per MAITE protocol.
- Moves to align with MAITE v0.7.0 by including _metadata_ attribute to estimators, augmentations, metrics and dataset definitions.
- Notebooks and tests updated based on MAITE alignment and fixes.
- Removes the constraint of models passed to JaticPyTorchObjectDetector being one of DETR, FasterRCNN and YOLO - thus facilitating other PyTorch detector models. (i.e. those implemented in torchvision)

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Remediation of vulnerabilities within `poetry.lock`
- Inclusion of `py.typed` for type checking support
- Integration of `pyright`

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Refactored individual object detectors to a single common class
- Included metadata indicating supported object detector models
- Refactored notebooks to demonstrate common detector usage

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Enable support for python 3.11

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Core HEART library aligned to MAITE v0.6.0
- attack, estimator and metrics accept Sequence of ArrayLike images
- utility function, process_inputs_for_art, processes Sequence of ArrayLike images
- Gradio app aligns with MAITE v0.6.0
- Notebooks updated to align with MAITE v0.6.0
- DPatch and Laser attacks tests updated to align with new data format

https://github.com/IBM/heart-library

https://pypi.org/project/heart-library/

Update Notes

- Alignment of HEART to MAITE v0.5.0
- MAITE v0.5.0 protocols: https://github.com/mit-ll-ai-technology/maite/blob/v0.5.0/examples/protocol_overview.ipynb
- Core library refactor (estimators, attacks, metrics)
- Refactor of all notebooks
- Refactor of Gradio app
- Added support for mAP metric
- Support for Feature Adversary attack
- Attack description: https://arxiv.org/abs/1511.05122
- ART implementation: https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/art/attacks/evasion/feature_adversaries/feature_adversaries_pytorch.py
- Demonstration added to notebook 3
- Support for black-box patch attack
- Attack description: https://arxiv.org/abs/1906.11897
- ART implementation: https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/art/attacks/evasion/dpatch_robust.py
- Demonstration added to notebook 6
- Support for Faster RCNN object detector
- Estimator description: https://arxiv.org/abs/1506.01497
- ART implementation: https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/art/estimators/object_detection/pytorch_faster_rcnn.py
- Demonstration added to notebook 6