ART 1.15.0

[beat-buesser]

This release of ART 1.15.0 introduces

Added

• Added default training function to TensorFlowV2Classifier (Implement Default Training Loop for TensorFlowV2Classifier #2124)
• Added TRADES adversarial training protocol in PyTorch (TRADES adversarial training protocol #2131)
• Added preprocessors for images supporting padding and resizing in PyTorch, TensorFlow and framework-independent (Implement Square Pad and Resize Preprocessors #2138)
• Added support for arbitrarily sized images in BadDet poisoning attacks (BadDet Poisoning Attacks Extension #2189)
• Added estimator for DEtection TRansformer (DETR) object detection models based on transformer architectures (Detection Transformer Estimator #2192)

Changed

• Changed PyTorch estimators to use PyTorch datasets and dataloaders to optimize the fit and predict methods for PyTorchClassifier, PyTorchRegressor, PyTorchRandomizedSmoothing, PyTorchObjectDetector, and PyTorchYolo and optimized the predict method of TensorFlowV2Classifier by using a TensorFlow dataset and applying @tf.function decorator (Optimize PyTorch Classifiers and Object Detectors #2180)
• Changed PyTorchObjectDetector to apply channels_first argument and improved performance by applying batch processing provided by newer PyTorch versions. (Optimize PyTorch Classifiers and Object Detectors #2180)

Removed

[None]

Fixed

• Fixed unnecessary duplicate prediction calls to estimator in SignOPTAttack (Remove incorrectly duplicated queries in SignOPT attack #2129)
• Fixed missing transfer of tensor to device in ProjectedGradientDescentPyTorch (Fix Tensor Device Inconsistencies in PGD #2135)
• Fixed trigger placement for image poisoning perturbations by correctly accessing height and width of the trigger image instead of swapping both (Fix Image Poisoning Perturbations Trigger Placement Bug #2143)
• Fixed key error in loss gradients of PyTorchYolo estimator and updated format of targets passed to the estimator in AdversarialPatchPyTorch to reflect updates to PyTorchYolo(Fix AdversarialPatchPyTorch compatibility with YOLO estimator #2169)
• Fixed Visible Deprecation Warning in analyze_by_distance and analyze_by_size of ClusteringAnalyzer (Fix ragged nested sequence warning #2195)

---

This discussion was created from the release ART 1.15.0.