Image registration/optical flow of 3D brain MRIs: fast calibration & transformation into standard space. Utilized convolutional neural networks on modern machine learning frameworks PyTorch and TensorFlow. First goal of the project is to avoid the optimization type operations typically related to calibrating images, greatly speeding up the computation speed.
Currently playing with some tensorflow code by written by Ignacio Rocco, from his the paper "Convolutional neural network architecture for geometric matching"