OpenPack
收藏资源简介:
OpenPack数据集是一个大规模、多模态的数据集,旨在全面捕捉现代物流中心真实包装操作的复杂性。该数据集在模拟环境中收集,包含53.8小时的传感器数据,记录了104次采集会话,有20,161个操作实例和53,286个动作实例。包装操作分为10个主要活动类别,细粒度的动作分为32个类别,详细描述了订单处理工作流程的每个步骤。该数据集通过深度学习方法(Attention Autoencoder和条件生成对抗网络)生成数据集,并通过数据洗牌增强数据集的多样性。
The OpenPack Dataset is a large-scale, multimodal dataset designed to comprehensively capture the complexity of real-world packaging operations in modern logistics centers. Collected in a simulated environment, the dataset contains 53.8 hours of sensor data, spanning 104 acquisition sessions, 20,161 operational instances and 53,286 action instances. Packaging operations are categorized into 10 major activity classes, while fine-grained actions are divided into 32 classes, detailing every step of the order processing workflow. This dataset was generated using deep learning methods including Attention Autoencoder and Conditional Generative Adversarial Network (CGAN), and its diversity was enhanced via data shuffling.

- 1SOS: A Shuffle Order Strategy for Data Augmentation in Industrial Human Activity Recognition越南国家大学胡志明市技术大学计算机科学与工程学院 · 2025年



