The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider
removing the
loading script
and relying on
automated data support
(you can use
convert_to_parquet
from the datasets library). If this is not possible, please
open a discussion
for direct help.
Part of MONSTER: https://arxiv.org/abs/2502.15122.
| PAMAP2 | |
|---|---|
| Category | HAR |
| Num. Examples | 38,856 |
| Num. Channels | 52 |
| Length | 100 |
| Sampling Freq. | 100 Hz |
| Num. Classes | 12 |
| License | Other |
| Citations | [1] |
PAMAP2 is a collection of data obtained from three Inertial Measurement Units (IMUs) placed on the wrist of the dominant arm, chest, and ankle, as well as 1 ECG heart rate [1]. The data was recorded at a frequency of 100Hz. The dataset includes annotated information about human activities performed by 9 subjects, each with their own unique physical characteristics. The majority of the subjects are male and have a dominant right hand. Notably, the dataset includes only one female subject (ID 102) and one left-handed subject (ID 108). In total, there are 12 different human activity classes represented in the dataset. The processed dataset contains 38,856 time series each of length 100 (i.e., representing one second of data per time series at 100 Hz). To ensure an unbiased evaluation, we divide the dataset into cross-validation folds based on the subjects.
[1] Attila Reiss and Didier Stricker. (2012). Introducing a new benchmarked dataset for activity monitoring. In 16th International Symposium on Wearable Computers, pages 108–109.
- Downloads last month
- 183