tony-wolff / au_bs_model

A machine learning model mapping action unit from OpenFace to Blendshapes from Arkit LiveFace

Machine Learning model mapping Action Units to arkit-compatible blendshapes

---

This repository contains model checkpoints and data for the mapping of OpenFace action units to blendshapes directly extracted with Arkit Liveface

• model.ipyng : the code, training and testing of the models
• pytorch models : folder containing multiple models
• training.md : file to track the performances of the models
• image_for_md : image folder for training.md
• au_bs_lagMeasure.ipynb : notebook that recorded the delay between action unit scripts and blendshape scripts
• AU26_r_JawOpen-lagMeasure.csv : file generated by the above notebook. Each au script is associated with its bs script, the framerate for both files, the lag between both files, and the number of frames for both files
• preprocessing_and_giga_dataset.ipynb : notebook to preprocess data (not available yet) and create the dataset used for my models
• outliers : folder containing plotting of Action Unit 26 VS JawOpen blendshape of scripts that have a too large lag to be included in the final dataset
• JawOpen_AU26_plots : folder of images containing the plot of Action Unit 26 VS JawOpen blendshape of all scripts

au_bs_model's Issues

Where is the demo.py file？