Pyrregular: Irregular Time Series Datasets and Benchmarks

Published at ICLR 2026 · 38 datasets, including the benchmark of 34 datasets and 12 classifiers · GitHub · Documentation · PyPI · Paper

Naturally irregular time series (uneven sampling, missing observations, signals recorded at different times, variable-length sequences) in one standardized format, for classification and regression. These are the datasets of the pyrregular Python library, including the 34 of the ICLR 2026 benchmark. The files are HDF5, read by the library (the dataset viewer is off):

pip install pyrregular

from pyrregular import load_dataset
df = load_dataset("Garment.h5")

The files are versioned with git tags (data-v1, ...); each pyrregular release downloads one fixed version.

Licenses

Each dataset keeps the license of its original source; we redistribute it in a converted format, with attribution. unknown means the source states no license (most of these come from the UEA & UCR time series archive). If you hold the rights to a dataset and want it removed or its attribution changed, please open an issue at https://github.com/fspinna/pyrregular/issues.

Dataset	License	Source	Citation
Abf	CC-BY-4.0	source	cite
Ais	CC-BY-4.0	source	cite
AllGestureWiimoteX	unknown	source	cite
AllGestureWiimoteY	unknown	source	cite
AllGestureWiimoteZ	unknown	source	cite
Animals	other	source	cite
AsphaltObstaclesCoordinates	unknown	source	cite
AsphaltPavementTypeCoordinates	unknown	source	cite
AsphaltRegularityCoordinates	unknown	source	cite
CharacterTrajectories	CC-BY-4.0	source	cite
CombinedTrajectories	CC-BY-NC-SA-4.0	source	cite
DodgerLoopDay	CC-BY-4.0	source	cite
DodgerLoopGame	CC-BY-4.0	source	cite
DodgerLoopWeekend	CC-BY-4.0	source	cite
Garment	CC-BY-4.0	source	cite
Geolife	unknown	source	cite
GeolifeSupervised	unknown	source	cite
GestureMidAirD1	unknown	source	cite
GestureMidAirD2	unknown	source	cite
GestureMidAirD3	unknown	source	cite
GesturePebbleZ1	unknown	source	cite
GesturePebbleZ2	unknown	source	cite
InsectWingbeat	unknown	source	cite
JapaneseVowels	CC-BY-4.0	source	cite
Ldfpa	CC-BY-4.0	source	cite
MelbournePedestrian	CC-BY-4.0	source	cite
Mimic3	ODbL-1.0	source	cite
PLAID	CC-BY-4.0	source	cite
Pamap2	CC-BY-4.0	source	cite
Physionet2012	ODC-By-1.0	source	cite
Physionet2019	CC-BY-4.0	source	cite
PickupGestureWiimoteZ	unknown	source	cite
Seabirds	CC0-1.0	source	cite
ShakeGestureWiimoteZ	unknown	source	cite
SpokenArabicDigits	CC-BY-4.0	source	cite
TDrive	unknown	source	cite
Taxi	CC-BY-4.0	source	cite
Vehicles	unknown	source	cite

Citation

Please cite the original source of each dataset (Citation column) and pyrregular:

@inproceedings{
    spinnato2026pyrregular,
    title={{PYRREGULAR}: A Unified Framework for Irregular Time Series, with Classification Benchmarks},
    author={Francesco Spinnato and Cristiano Landi},
    booktitle={The Fourteenth International Conference on Learning Representations},
    year={2026},
    url={https://openreview.net/forum?id=qetBM8nLkf}
}