English | 简体中文
AdvBox is a series of AI model security tools set of Baidu Open Source,including the generation, detection and protection of adversarial examples, as well as attack and defense cases for different AI applications. It contains a toolbox to generate adversarial examples that fool neural networks in Paddle, PyTorch, etc. The existence of adversarial examples may be an inherent weakness of deep learning models. They can be used to train networks or benchmark the robustness.
Robustness is a robustness benchmark tool for vision DNN models. It supports image classification models on PaddlePaddle, PyTorch, etc. Robustness is Equally Important as Accuracy for the Real-World, Safety-critical scenarios. We can use Robustness's command line tool to generate noise for quantifying DNN model robustness to the Real-World threats.
PaddleSleeve is provided under the Apache-2.0 license.