dylanorange/geal
收藏Hugging Face2024-12-12 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/dylanorange/geal
下载链接
链接失效反馈官方服务：
资源简介：
GEAL框架旨在通过利用预训练的2D模型来增强3D affordance学习的泛化性和鲁棒性。为了促进跨多样现实场景的鲁棒3D affordance学习，作者基于PIAD和LASO数据集的测试集建立了两个3D affordance鲁棒性基准：PIAD-C和LASO-C。这些基准应用了七种类型的干扰，每种干扰有五个严重级别，总共包含4890个对象-affordance配对，涉及17个affordance类别和23个对象类别，以及2047个不同的对象形状。
GEAL is a novel framework designed to enhance the generalization and robustness of 3D affordance learning by leveraging pre-trained 2D models. The dataset includes two 3D affordance robustness benchmarks: PIAD-C and LASO-C, based on the test sets of the commonly used datasets PIAD and LASO. The dataset applies seven types of corruptions, each with five severity levels, resulting in a total of 4890 object-affordance pairings, comprising 17 affordance categories and 23 object categories with 2047 distinct object shapes.
提供机构：
dylanorange
