Cross-Domain Transferability of Adversarial Perturbations

Authors: Muhammad Muzammal Naseer, Salman H. Khan, Muhammad Haris Khan, Fahad Shahbaz Khan, Fatih Porikli

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments
Researcher Affiliation Collaboration 1Australian National University, Canberra, Australia 2Inception Institute of Artificial Intelligence, Abu Dhabi, UAE 3CVL, Department of Electrical Engineering, Linköping University, Sweden
Pseudocode Yes The overall training scheme for the generative network is given in Algorithm 1.
Open Source Code Yes Code is available at: https://github.com/Muzammal-Naseer/Cross-domain-perturbations
Open Datasets Yes Datasets. We consider the following datasets for generator training namely Paintings [29], Comics [30], Image Net and a subset of Chest X-ray (Chest X) [28].
Dataset Splits Yes Inference: Inference is performed on Image Net validation set (val-set) (50k samples), a subset (5k samples) of Image Net proposed by [11] and Image Net-Neur IPS [31] (1k samples) dataset.
Hardware Specification No The paper describes the model architecture, optimizer, and training details, but does not specify the hardware (e.g., GPU models, CPU, memory) used for running the experiments.
Software Dependencies No The paper mentions using a ResNet architecture and Adam optimizer, but does not specify the versions of software libraries or frameworks (e.g., TensorFlow, PyTorch) used for implementation.
Experiment Setup Yes For training, we used Adam optimizer [23] with a learning rate of 1e-4 and values of exponential decay rate for first and second moments set to 0.5 and 0.999, respectively.