KIFNet: Continuous Prediction of Leg Kinematics during Walking using Inertial Sensors, Smart Glasses, and Embedded Computing

This is a repository, containing training source code we used for our paper accepted to ICRA 2023.

In order to run our code you will need:

1. Egocentric Vision & Kinematics dataset. Download it and change appropriate data config files to match your paths.
2. ml-mobileone package and model weights (we used S0 unfused), that you need to place in ./ml-mobileone/weights/ folder.
3. An existing Weight & Biases project. We used W&B for experiment tracking and configuration, hence it is required to run our pipeline if you are not willing to modify the code.

Contributors:

• Oleksii Tsepa
• Roman Burakov