AnySkin Slip Detection Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://any-skin.github.io/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集旨在训练滑动检测模型，使用了AnySkin传感器收集的触觉信号，涵盖了各种日常物品，并对其是否滑动进行了相应的标注。该数据集包含了在不同抓握条件下收集的、被标记为滑动或未滑动的序列，使得模型在未见过的物品上达到了92%的准确率。数据集规模涉及40种日常物品（其中30种用于训练，10种用于评估），任务目标是滑动检测。
This dataset is designed for training slip detection models. It employs tactile signals collected by AnySkin sensors, covers a wide range of daily objects, and is correspondingly annotated with slip status. The dataset contains sequences collected under diverse grasping conditions, labeled as either slipped or non-slipped, enabling the model to reach 92% accuracy on unseen objects. In total, the dataset involves 40 types of daily objects, with 30 used for training and 10 for evaluation, and the task objective is slip detection.
提供机构：
AnySkin research team
