Screenshot of a simulated physical model using Projective Dynamics. The user can freely change the distributions of muscles and fat. My next task was to find the muscle and fat parameters that can best fit the X-Ray silhouettes.
My latest research at Movement Research Lab was to generate a personalized 3D musculoskeletal model from EOS X-Ray images, which can be used for visualization and simulation.
My approach for this research was to use ideas in physical simulation and differentiable rendering. I used Projective Dynamics to create a quasistatic simulation where the user can change the muscle and fat distribution of the musculoskeleton. After finishing the modelling, my goal was to devise an algorithm that can fit muscle and fat parameters to the silhouettes of two X-Ray images (front and side).
Other than that, I was in the Biomechanics Project team at Movement Research Lab, I have done the following work:
Cleanup and convert C3D gait motion data (obtained from Seoul National Bundang Hospital) to BVH files
Research on unsupervised learning of gait data using VAEs.
Wrote code for porting MASS (Muscle-Actuated Skeletal System) on a distributed cluster, using Ray and RLlib.
Wrote code for analyzing gait properties (metabolic energy expenditure, joint torques, muscle activation levels) for a musculoskeletal system with exoskeleton device attached.
Miscellaneous work related to paper submission (experiments, analysis, writing)