A part of UQ's Winter Research Program 2021, Machine Learning of Biological Swimmers. Undertaken from 21-Jun-2021 to 23-Jul-2021, supervised by Dr Alex Stilgoe.
Using MATLAB under attic/, synthetic images of particles are generated alongside a labelled dataset. An example run of the MATLAB files are under data.zip.
Three single-output CNNs based off DeepTrack 2.0's framework were used to predict the angle and x- and y-coordinates of particles.
Sample Image Output. Predicted (red) against gold standard (yellow) coordinate and angles.
short_animation.mp4
Short Animation of Model Predictions (red) against gold standard (yellow) for a simple walk.