ART 1.7.1
This release of ART 1.7.1 provides updates to ART 1.7.
Added
- Added wrapper
Mp3CompressionPyTorchforMp3Compressionto make it compatible with PyTorch-specific attack implementations. (#1210) - Added new install option
non-frameworktosetup.pyto install all non-framework dependencies of ART. (#1209) - Added wrapper
VideoCompressionPyTorchforVideoCompressionto make it compatible with PyTorch-specific attack implementations. (#1210)
Changed
- Changed
Mp3Compressionto add back reapplication of normalization to the compressed result. (#1210) - Changed
KerasClassifier.fitto use batching provided by the methodfitof the Keras model. (#1182)
Removed
[None]
Fixed
- Fixed bug of not passing user-provided device type, and instead always using default
gpu, to standardisation preprocessor in allPyTorchEstimatorby using user-provided device type. (#1223) - Fixed bug in method
BaseEstimator.fit_generatorfor fitting generators in cases where preprocessing is defined to not apply preprocessing twice. (#1219) - Fixed bug in
ImperceptibleASRPyTorchto prevent NaN loss value for batch size larger than 1 by removing unnecessary zero-padding. (#1198) - Fixed two bugs in
OverTheAirFlickeringPyTorchby making sure that the regularization norms are computed over the whole batch of perturbations, rather than per sample's perturbation and second that the "roll" operations are performed over the batch samples, rather than over the frames. (#1192) - Fixed bug in
SpectralSignatureDefense, that lead to rejections of all clean images, by correctly indexing the label data. (#1189) - Fixed bug of accidentally removed checks for
apply_fitandapply_predictproperties of framework-independentPreprocessortools inPyTorchEstimatorandTensorFlowV2Estimator. With the bug thePreprocessortools were always applied in methodsfitandpredictindependent of the values ofapply_fitandapply_predict. (#1181) - Fixed bug in
MembershipInferenceBlackBoxRemove.inferby removing unnecessary shuffling of the test data. (#1173) - Fixed bug in
PixelAttackandThresholdAttackby casting input data to correct dtype. (#1175)