PartImageNet++
收藏资源简介:
PartImageNet++是由清华大学等机构创建的一个大型数据集,专门为ImageNet-1K的所有类别提供高质量的部分分割注释。该数据集包含100,000张图像,涵盖1,000个对象类别和3,310个部分类别,旨在通过详细的部分注释方案增强模型的鲁棒性。数据集的创建过程包括手动注释和质量控制,确保注释的高质量。PartImageNet++主要应用于提高对象识别系统的鲁棒性,特别是在对抗性扰动和常见图像损坏的情况下。
PartImageNet++ is a large-scale dataset created by Tsinghua University and other institutions, which specifically provides high-quality part segmentation annotations for all categories in ImageNet-1K. This dataset contains 100,000 images, covering 1,000 object categories and 3,310 part categories, aiming to enhance the robustness of models through a detailed part annotation scheme. The creation process of the dataset includes manual annotation and quality control to ensure the high quality of annotations. PartImageNet++ is mainly applied to improving the robustness of object recognition systems, especially under adversarial perturbations and common image corruptions.
PartImageNetPP
数据集概述
- 名称:PartImageNetPP
- 来源:论文 "PartImageNet++ Dataset: Scaling up Part-based Models for Robust Recognition" (ECCV 2024)
- 可用性:当前数据集可在 https://huggingface.co/datasets/lixiao20/PartImageNetPP 获取
引用信息
@inproceedings{li2024pinpp, author = {Li, Xiao and Liu, Yining and Dong, Na and Qin, Sitian and Hu, Xiaolin}, title = {PartImageNet++ Dataset: Scaling up Part-based Models for Robust Recognition}, booktitle={European conference on computer vision}, year = {2024}, organization={Springer} }




