L-AVATeD: The Lidar and Visual Walking Terrain Dataset
Author Information
Author(s): David Whipps, Patrick Ippersiel, Philippe C. Dixon
Primary Institution: Université de Montréal
Conclusion
The L-AVATeD dataset provides a novel, open-source collection of visual and LiDAR image pairs for human walking terrain, which can enhance the development of control systems for exoskeletons and gait analysis.
Supporting Evidence
- The dataset includes 7,968 RGB/LiDAR image pairs.
- It captures nine different terrain types relevant to walking environments.
- The data collection was conducted over six months to ensure diversity in terrain samples.
- The dataset is open-source, allowing for broad access and use in research.
Takeaway
This study created a big collection of pictures and data about different walking surfaces to help robots and researchers understand how to walk better on various terrains.
Methodology
Data were collected using iPhones equipped with LiDAR sensors to capture RGB and depth images of various walking terrains over six months.
Potential Biases
Potential bias in terrain classification accuracy due to varying patient pathologies affecting IMU data.
Limitations
The dataset has class imbalance in terrain types, which may affect the training of models.
Participant Demographics
Three young healthy adult members of the research lab participated in data collection.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website