Janus: Multi-Task Head Pose Recognition Model

Model Description

Janus is a multi-class head pose recognition model that identifies key directional movements of the head. Named after the Roman god with two faces, Janus symbolizes different directions and perspectives. This model is designed for real-time applications such as gesture recognition, virtual meetings, and interactive systems.

Key Features

  • Framework: Keras
  • Model Type: Functional (with combined multi-task outputs)
  • Input: 468 facial landmarks (x, y, z coordinates) from MediaPipe
  • Output: Classifications for head direction, motion, and up/down movement
  • Tasks:
    • Head Direction: Front, Left, Right
    • Head Motion: Tilt, Turn
    • Up/Down Motion: Up, Down, Unknown

Limitations

  • May not perform well on extreme or non-standard head poses
  • Optimized primarily for front-facing camera angles

Ethical Considerations

  • Use responsibly and avoid applications that could misuse facial data
  • Respect privacy and data protection regulations

License

This model is released under the MIT License.

Citation

If you use this model in your research, please cite:

@inproceedings{SUKO KUO_2024_janus,
  title={Janus: Multi-Task Head Pose Recognition Model},
  author={SUKO KUO},
  year={2024},
  url={https://huggingface.co./suko/Janus}
}
Downloads last month
0
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Evaluation results