OpenPack: Public multi-modal dataset for packaging work recognition in logistics domain

OpenPack is an open access logistics-dataset for human activity recognition, which contains human movement and package information from ten subjects. The package information includes the size and number of items included in each packaging task. While the human movement information is subdivided into three types of data, acceleration, physiological, and depth-sensing.

In the "Humanware laboratory" at IST Osaka University, with the supervision of industrial engineers, an experiment to mimic logistic center labor was designed. Workers with previous packaging experience performed a set of packaging tasks according to an instruction manual from a real-life logistic center. During the experiment, subjects were recorded while performing packing operations using Lidar, Kinect, and Realsense depth sensors while also wearing four IMU devices and 2 Empatica E4 wearable sensors. Each subject was tasked with performing 20 packing jobs in 5 separate sessions for a total of 100 jobs. Approximately 50 hours of packaging operations have been labeled into 10 global operation classes and 16 action classes for this dataset.

We are hosting an activity recognition competition, using this dataset (OpenPack v0.3.x) at a PerCom 2023 Workshop! The task is very simple: Recognize 10 work operations from the OpenPack dataset. Please visit our website and check the details. https://open-pack.github.io/challenge2022

Work is continuously being done to update and improve this dataset. When downloading and using this dataset please verify that the version is up to date with the latest release. The latest release [0.3.0] was uploaded on 01/10/2022. You can find information on how to use this dataset at https://github.com/open-pack/openpack-toolkit.