TULIP Dataset - TULIPv1 [2024]

A description of my image.

Dataset Configuration

  • Subjects: 15 subjects (Ten were clinically diagnosed with PD, while five had no prior PD diagnosis.)
  • Cameras: 6 cameras (1920 × 1200 pixels; 80 fps; Basler acA1920-155uc) simultaneous recording
  • Arena: 6.3 × 3 meter hexagonal space
  • Activities: 25 activities (Part III: Motor Examination UPDRS components. With clinician guidance, we recorded 21 activities from which UPDRS scores could be derived, including unilateral tasks, such as guided hand movement, index finger tapping, stepping, and bilateral tasks, such as arm straightening, gait. Four additional unilateral activities, including tasks like standing on one leg and finger tapping using all fingers, were incorporated to encompass a wider range of behavioral features pertinent to UPDRS scoring)
  • Videos: MP4 format, sliced to include each activity for every subject.
  • Labels: 29 UPDRS scores (0 to 4) from three clinical experts from the videos

Activities & Guidelines

A description of my image.

Relationship between UPDRS criteria and TULIP dataset

A description of my image. A description of my image.

How to get data

To access the data, please submit a request via this Google Form. Currently, the dataset can only be shared for academic research at institutions with a human subjects Institutional Review Board (IRB). This process is necessary due to the sensitive nature of the dataset. We take the privacy and security of the data seriously and will review each request to ensure responsible data handling and compliance.

We are in the process of releasing a de-identified version of the dataset. The de-identified dataset will also be made available only for academic research purposes, but IRB agreement requirements will be relaxed. If you would like to be notified when the de-identified dataset is available, please fill out the Google form and note that you would like to be added to the de-identification notification list.

Question

If you have any questions about the data and the paper, please contact kyungdo.kim@duke.edu .

Citation

@InProceedings{Kim_2024_CVPR,
    author    = {Kim, Kyungdo and Lyu, Sihan and Mantri, Sneha and Dunn, Timothy W.},
    title     = {TULIP: Multi-camera 3D Precision Assessment of Parkinson's Disease},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2024},
    pages     = {22551-22562}
}

TULIP Project

Duke White