ARES-Bench
收藏arXiv2023-02-28 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2302.14301v1
下载链接
链接失效反馈官方服务：
资源简介：
本文建立了一个名为ARES-Bench的综合鲁棒性基准，用于图像分类任务。该基准评估了55种典型深度学习模型在ImageNet上的鲁棒性，这些模型具有多样的架构（如CNNs, Transformers）和学习算法（如正常监督训练、预训练、对抗训练），并受到多种对抗攻击和分布外（OOD）数据集的影响。
This paper establishes a comprehensive robustness benchmark named ARES-Bench for image classification tasks. This benchmark evaluates the robustness of 55 representative deep learning models on ImageNet, which feature diverse architectures (e.g., CNNs, Transformers) and learning algorithms (e.g., standard supervised training, pretraining, adversarial training), and assesses their performance under various adversarial attacks and out-of-distribution (OOD) datasets.
创建时间：
2023-02-28
