Dataset Card for SIMSPINE

Dataset Summary

SIMSPINE is a dataset of additional 3D spine keypoint annotations derived from the Human3.6M dataset. The dataset provides 15 simulated spine-related joints aligned with Human3.6M motion capture sequences.

The annotations are generated using a biomechanical simulation pipeline built on OpenSim, combined with pseudo-labeling derived from the original Human3.6M keypoints and video frames.

SIMSPINE extends the standard Human3.6M skeleton by providing additional spine-related joints that enable:

• fine-grained modeling of spinal motion
• biomechanical analysis
• improved human pose estimation for spine-aware models
• evaluation of spine reconstruction algorithms

The dataset does not include any original Human3.6M data.

Instead, it provides additional annotations aligned with Human3.6M frames, which can be combined with the original dataset after users obtain it independently.

Supported Tasks

SIMSPINE can be used for:

• 2D human pose estimation
• 3D human pose estimation
• spine pose estimation
• human motion analysis
• biomechanical modeling
• 2D-to-3D pose lifting research
• human pose reconstruction

Dataset Structure

TBA.

Dataset Creation

Source Dataset

SIMSPINE is derived from the Human3.6M dataset.

Human3.6M contains:

• synchronized multi-view videos
• motion capture ground truth
• 3D human joint annotations (17 joints)

SIMSPINE does not redistribute any of this data.

Users must obtain Human3.6M independently from the official source: https://vision.imar.ro/human3.6m/

Annotation Pipeline

The SIMSPINE annotations were generated using the following pipeline:

1. Human3.6M keypoints and videos were used to estimate additional anatomical landmarks.

2. Pseudo-labels were generated for intermediate anatomical points along the torso using 2D detectors from SpinePose.

3. A biomechanical spine model was constructed using OpenSim.

4. The pseudo-labels were used to simulate 15 anatomically plausible spine-related joints.

5. The resulting joints were aligned with the Human3.6M coordinate system and frames.

The final dataset therefore consists of simulated spine joints consistent with the original Human3.6M motion sequences.

Dataset Statistics

The dataset spans the same subjects, actions, and frames as Human3.6M.

More details TBA.

Intended Uses

SIMSPINE is intended for academic research in human pose estimation and biomechanics.

Typical use cases include:

• training spine-aware pose estimation models
• evaluating spine reconstruction methods
• studying spinal motion patterns
• improving full-body pose estimation

Out-of-Scope Uses

The dataset should not be used for:

• clinical diagnosis
• medical decision making
• commercial products without permission

Licensing Information

SIMSPINE is released under the SIMSPINE Academic Research License.

Key conditions:

• Academic use only
• No redistribution of the dataset
• Proper citation required

SIMSPINE is a derived dataset from Human3.6M.

Users must also comply with the Human3.6M license agreement, available at: https://vision.imar.ro/human3.6m/

SIMSPINE does not include:

• Human3.6M images
• Human3.6M videos
• Human3.6M motion capture data
• Human3.6M joint annotations

Users must independently obtain Human3.6M to use this dataset.

Citation

If you use SIMSPINE, please cite our paper:

@inproceedings{khan2026simspine,
  author = {Khan, Muhammad Saif Ullah and Stricker, Didier},
  title = {SIMSPINE: A Biomechanics-Aware Simulation Framework for 3D Spine Motion Annotation and Benchmarking},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2026},
}

You must also cite the Human3.6M papers:

@article{h36m_pami,
  author = {Ionescu, Catalin and Papava, Dragos and Olaru, Vlad and Sminchisescu, Cristian},
  title = {Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year = {2014}
}

@inproceedings{IonescuSminchisescu11,
  author = {Catalin Ionescu, Fuxin Li, Cristian Sminchisescu},
  title = {Latent Structured Models for Human Pose Estimation},
  booktitle = {International Conference on Computer Vision},
  year = {2011}
}

Acknowledgements

This dataset builds upon the Human3.6M dataset created by:

Catalin Ionescu, Dragos Papava, Vlad Olaru, and Cristian Sminchisescu.

Contact

For questions regarding SIMSPINE:

Muhammad Saif Ullah Khan Augmented Vision Group German Research Center for Artificial Intelligence (DFKI) Kaiserslautern, Germany

Email: muhammad_saif_ullah.khan@dfki.de