Farhan Ahmed

Implement Cutout in Numpy and PyTorch

4

# Description Implementation of the Cutout data augmentation defense in the Numpy and PyTorch frameworks. Fixes # (issue) ## Type of change Please check all relevant options. - [ ]...

Implementation of DP-InstaHide and Data Augmentation Defenses

1

**Is your feature request related to a problem? Please describe.** Mixup, CutOut, CutMix, and Maxup are relatively simple data augmentation defenses that have been shown to provide robustness against poisoning...

Subclass `PyTorchYolo` and `PyTorchDetectionTransformer` off `PyTorchObjectDetector`

2

# Description Modify both the `PyTorchYolo` and `PyTorchDetectionTransformer` estimators to be subclassed off the `PyTorchObjectDetector` estimator. This reduces a lot of redundant code and allows most functionality to be shared...

Hugging Face Language Models Implementation

2

# Description Implementation of language models under a new `art.estimators.language_modeling` submodule. Currently only Hugging Face language models using a PyTorch back-end have been implemented. This is implemented as the `HuggingFaceLanguageModel`...

**Is your feature request related to a problem? Please describe.** Previously, the `HuggingFaceClassifierPyTorch` was implemented for the classification task. A similar estimator should be implemented for the object detection task....

YOLO Object Detection Estimator for TensorFlow

1

**Is your feature request related to a problem? Please describe.** The current YOLO object detection estimator only has an implementation for PyTorch. There is no YOLO estimator for TensorFlow. **Describe...

**Is your feature request related to a problem? Please describe.** In the original paper for the Witches' Brew / Gradient Matching Attack (), the two names were originally interchangeable. Therefore,...

**Is your feature request related to a problem? Please describe.** Under `art.attacks.poisoning`, although the `PoisoningAttackBackdoor` object exists which allows the user to insert backdoors and perform the Dirty Label Backdoor...

Implement `get_activations()` for PyTorch Object Detectors

1

**Is your feature request related to a problem? Please describe.** The current PyTorch object detectors (`PyTorchObjectDetector`, `PyTorchFasterRCNN`, `PyTorchYolo`, `PyTorchDetectionTransformer`) do not implement the `get_activations()` method and therefore cannot extract activations....

Subclass the `PyTorchYolo` and `PyTorchDetectionTransformer` estimators off the `PyTorchObjectDetector` estimator class

1

**Is your feature request related to a problem? Please describe.** There is a lot of code overlap between the `PyTorchYolo`, `PyTorchDetectionTransformer`, and `PyTorchObjectDetector` estimators. This causes bugs to have to...