← All datasets

inhard Open

InHARD: Industrial Human Action Recognition Dataset

Dallel M., Havard V., Baudry D., Savatier X.

CESI LINEACT / ESIGELEC-IRSEEM · 2020 · France

RGB+skeleton dataset for industrial human action recognition in a collaborative-robotics context: 16 subjects performing assembly actions alongside a robot, labeled into 13 action and 74 meta-action classes. Skeleton data from a Perception Neuron suit at 120 Hz plus three synchronized RGB views. Presented at IEEE ICHMS 2020 (doi:10.1109/ICHMS49158.2020.9209531).

Optical mocapVideoAssembly
View source ↗DOI: 10.5281/zenodo.4003541

Details

Subjects
n = 16
Protocol / load
Industrial assembly actions in a human-robot collaboration cell; 13 action classes, 4,800+ action samples, 2M+ frames
Equipment
Perception Neuron 32 inertial mocap suit (17 joints, BVH skeletons); three Logitech C920 RGB cameras (two side views at ±45°, one top view)
Sampling
mocap 120 Hz
Formats
mp4, csv
License
CC-BY-4.0
Keywords
activity-recognition, har, human-robot-collaboration, assembly, skeleton, manufacturing
Added to catalog
2026-08-25

Cite

Cite the dataset as its authors request (check the source page for an official citation); this BibTeX is a convenience starting point.

@misc{inhard,
  author = {Dallel, M. and Havard, V. and Baudry, D. and Savatier, X.},
  title  = {{InHARD: Industrial Human Action Recognition Dataset}},
  year   = {2020},
  doi    = {10.5281/zenodo.4003541},
  url    = {https://doi.org/10.5281/zenodo.4003541},
}