2D Video-Analysis Technology to Analyze Gait
Individuals from the Restore Center, in collaboration with Gillette Children’s Specialty Healthcare, have developed a deep neural network model to predict common quantitative gait metrics, such as cadence, walking speed, and the gait deviation index (GDI), from a single-camera video. The model demonstrates good predictive accuracy when applied to videos of children diagnosed with cerebral palsy and compared against quantities from optical motion capture.
The study (https://www.nature.com/articles/s4146...) was led by Restore Center research associate Lukasz Kidzinki and published in Nature Communication on August 13, 2020.
Kidziński, Ł., Yang, B., Hicks, J.L. et al. Deep neural networks enable quantitative movement analysis using single-camera videos. Nat Commun 11, 4054 (2020). https://doi.org/10.1038/s41467-020-17...
Additional resources
Demo software: http://gaitlab.stanford.edu/
Access code: https://github.com/stanfordnmbl/mobil...
Download dataset of trajectories of landmarks extracted from videos: https://simtk.org/projects/video-gaitlab