Summary
This case study relates to our experience of releasing open-source data to the public domain.
Context
Motion analysis traditionally requires small reflective markers to be placed all over the body to enable the tracking of skeletal movement. Within the Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA) we have been developing and validating markerless approaches that aim to extract similar motion data from images alone. As part of this validation work, we synchronously collected marker-based and image data of 15 people performing a variety of movements (running, walking, jumping and hopping), through which we showed great promise in the markerless techniques to enable movement analyses (‘pose estimation’) outside of the laboratory, without any need to place anything on the participant.
As the field of pose estimation is incredibly fast moving, we set out to make this (‘bioCV’) dataset publicly available to enable other researchers to develop, train and validate new algorithms against our benchmark data.
Reflections
Were there any challenges or barriers to engaging in open research?
The main challenges we faced included:
Consent of individuals – ensuring participants were aware and consented to the release of identifiable information (video cannot be anonymised).
Data security/storage – high-definition video takes up a huge amount of storage, which must be organised very well.
Data security and storage was key – a thorough data management plan essential.
Ensuring data would be used appropriately in the future – for research purposes only.
Which resources or contacts did you find helpful?
The Research Data Service Team were incredibly useful with the organisation of the data set, curating the data set including hosting the download webpage and ensuring that external researchers agreed to the correct terms and conditions of the data use.
How has open research impacted your project?
Making the open source data set has opened up opportunities for broader collaboration with external researchers and promoted reproducibility in science.
What are the lessons learned?
Planning for data release early is essential - at the very inception of the study (ethics, data management plan and how/where the data will be hosted).
Take home message?
Open Science is the future!
About the author
Steffi Colyer is a Senior Lecturer in Biomechanics and co-leads the Bath Beacon: Sport & Technology in a Digital Society. Steffi is particularly interested in how we can use unobtrusive technology to monitor athletes.