AnySkin Slip Detection Dataset

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.