RobustART

RobustART is a comprehensive Robustness investigation benchmark on ImageNet regarding ARchitectural design (49 human-designed off-the-shelf architectures and neural architecture searched networks) and Training techniques (10+ general ones e.g., extra training data, etc) towards diverse noises (adversarial, natural, and system noises).

The benchmark (including open-source toolkit, pre-trained model zoo, datasets, and analyses):

• presents an open-source platform for conducting comprehensive evaluation on diverse robustness types;
• provides a variety of pre-trained models with different training techniques to facilitate robustness evaluation;
• proposes a new view to better understand the mechanism towards designing robust DNN architectures, backed up by the analysis.

For more details, please refer to the website. We will continuously contribute to building this ecosystem for the community.