RecGym Dataset: Gym Workouts Recognition Dataset with IMU and Capacitive Sensor

The RecGym dataset is a collection of gym workouts with IMU and Capacitive sensors, designed for research and development in recommendation systems and fitness applications.

The data set records ten volunteers' gym sessions with a sensing unit composed of an IMU sensor (columns of A_x, A_y, A_z, G_x, G_y, G_z) and a Body Capacitance sensor (column of C_1). The sensing units were worn at three positions: on the wrist, in the pocket, and on the calf, with a sampling rate of 20 Hz. The data set contains the motion signals of twelve activities, including eleven workouts: Adductor, ArmCurl, BenchPress, LegCurl, LegPress, Riding, RopeSkipping, Running, Squat, StairsClimber, Walking, and a "Null" activity when the volunteer hangs around between different workouts session. Each participant performed the above-listed workouts for five sessions in five days (each session lasts around one hour). Altogether, fifty sessions of normalized gym workout data are presented in this data set.

Dataset Overview

• Number of Exercises: 12 (Adductor, ArmCurl, BenchPress, LegCurl, LegPress, Riding, RopeSkipping, Running, Squat, StairsClimber, Walking, and a "Null" activity)
• Number of Sensors: 2 (IMU and HBC(Human Body Capacitance) sensor)
• Number of Sensors Positions: 3 (wrist, calf, pocket)
• Number of Features: 7 (Acc_xyz, Gyro_xyz and HBC)
• Number of Subjects: 10
• Number of Sessions: 10 x 5
• Number of Recording Hours: 10 x 5
• Purpose: To provide a robust dataset for building and testing human activity recognition systems in fitness contexts.

Access the Dataset

You can download the RecGym dataset from Kaggle: RecGym Dataset on Kaggle.

Data Loader

The dataloader of RecGym is available here for ease-of-use: RecGym Data Loader on Kaggle.

More Details

For more details, please visit:

• GitHub Repository for the dataset and applied sensing modalities
• Our paper: The Contribution of Human Body Capacitance/Body-Area Electric Field To Individual and Collaborative Activity Recognition, available at arXiv.

License

The dataset is available under the Creative Commons Attribution 4.0 International License.

Contact

For questions or feedback, please contact the dataset creator on Kaggle.