#This file contains the code corresponding to the article "Machine Learning to Map Inertial Motion Capture Inputs to Optical Motion Capture-driven Musculoskeletal Model Outputs".
#A HPC cluster was used to run the complete cross-validation runs which are not feasible to run on a normap laptop. However, few instances, can be easily done using the function "specific" and "specific_CV" in pytorch.py, respectively.
#Lastly, all the plots and statistics provided in the paper can be reproduced by simply running main.py. Note that the path need to be modified in main.py.
#Test data for reproducing the results are provided here. The full training dataset can be obtained on request to vikranth.harthikotenagaraja@eng.ox.ac.uk
#The codes are developed using Python 3 with keras for ML. #MSK_ML_beta #This file contains the code description used in the article: "Machine Learning for Optical Motion Capture-driven Musculoskeletal Modeling from Inertial Motion Capture Data" #The codes are developed in python3 using standard python modules and keras for ML.
#The code here contains:
#a) Pipeline to run ML methods such FFNN, RNN, and LM which can be easily extended for other methods.
#b) We have used this pipeline to run cross-validation on cluster.
#c) Final training and testing can be easily done on any laptop/computer.
#d) Code for analysis and plotting is also provided.
#Code description #pytorch_utilities.py -- contains function for various models (Linear, Neural Network, RNN, LSTM, GRU,....) and generate a file with hyperparameters choices #read_in_out.py -- contains classes to read input and output data and some other classes for handling the final trained models #pytorch.py -- several functions to handle data, perform cross-validation, train model, forward pass, plot and analyse results, plot outputs #find_best_model.py -- code for finding the best-fit model using the average validation accuracy. #main.py -- contains function to estimate plot results and compute statistics
#final trained NN model #Files are heavy to upload, can be downloaded from here: https://www.dropbox.com/sh/svuqdy597d6pg60/AACiWr-kVx_W0bU0AD3HWPNha?dl=0 #Test data for reproducing the results are provided here, Input data -- https://www.dropbox.com/sh/8isp6yl29np6ngo/AAAqRWIc8lTOtYieehUSEQN1a?dl=0 Output data -- https://www.dropbox.com/sh/7h1oncpyru9vupl/AACm0YPmlmdu2rlabUpynCW4a?dl=0 The full training can be obtained on request to vikranth.harthikotenagaraja@eng.ox.ac.uk