University of Technology Chemnitz, Germany
Department Robotics and Human Machine Interaction
Author: Robert Schulz

TUC-AR Dtaset Card

TUC-AR is a small scale action recognition dataset, containing 6(+1) action categories for human machine interaction. This version contains video sequences, stored as images, frame by frame.

Dataset Details

• RGB and depth input recorded by Intel RealSense D435 depth camera
• 8 subjects
• 11,031 sequences (train 8,893/ val 2,138)
• 3 perspectives per scene
• 6(+1) action classes

  Action	Label
  A000	None
  A001	Waving
  A002	Pointing
  A003	Clapping
  A004	Follow
  A005	Walking
  A006	Stop

Use

1. Install the RSProduction Machine Learning package (PyPi, GitHub)

pip install rsp-ml

2. Use the HF datasat with rsp.ml.dataset.TUC_AR

from rsp.ml.dataset import TUC_AR
import rsp.ml.multi_transforms as multi_transforms

transforms = multi_transforms.Compose([multi_transforms.Resize((400, 400))])
tuc_ar_ds = TUC_AR(
              split='val',
              depth_channel=True,
              transforms=transforms,
              num_actions=10,
              streaming=True)

Dataset Card Contact

In case of any doubts about the dataset preprocessing and preparation, please contact TUC RHMi.

Acknowledgement

Thanks go out to Flower Labs. The structure and layout of this HuggigFace dataset is inspired by their dataset flwrlabs/ucf101.