This dataset is created for better understandind human behaviors in assembly tasks, with a focus on the upper limb motion.
This data includes 5 tasks. Each task requires the human to use LEGO pieces to assemble an object. The goal objects are different for different tasks, but they may share common LEGO pieces. For each trial, the human subject was given one of the following goal images. The human then used the component pieces on the table to assemble the goal object based on his/her understanding.
Goal objects for task 1-5.
Click for the video.
The motion data was recorded using the Intel RealSense D415. The human motion was extracted from the raw data using the OpenPose.
The data focuses on the human arm motion throughout the tasks, including the shoulders, elbows, and wrists. The data is stored as trajectories of the joints. The trajectory profiles are in the form of
- Tx18: [Left Wrist XYZ, Right Wrist XYZ, Left Elbow XYZ, Right Elbow XYZ, Left Shoulder XYZ, Right Shoulder XYZ].
T is the total timesteps in a trial.