niuniandaji/mini-imagenet-c
收藏Hugging Face2025-09-27 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/niuniandaji/mini-imagenet-c
下载链接
链接失效反馈官方服务：
资源简介：
MiniImageNet-C是一个面向计算机视觉模型鲁棒性测试的紧凑型ImageNet-C数据集，包含15种不同类型的图像腐蚀，每个类别下每种腐蚀类型包含50个图像，共覆盖1000个类别，总计750,000张图像，所有图像均为腐蚀程度最高的第5级。该数据集仅包含一个测试集，用于快速鲁棒性评估、降低计算需求以及教育和原型设计。
MiniImageNet-C is a compact version of the ImageNet-C dataset designed for testing the robustness of computer vision models against various image corruptions. It includes 15 different types of corrupted images, with 50 images per category per corruption type, across 1000 categories, totaling 750,000 images, all at the most severe level of corruption (level 5). The dataset consists of a single test split, intended for quick robustness evaluation, reduced computational requirements, and educational and prototyping purposes.
提供机构：
niuniandaji
