README

Supplementary data for the paper:
Data-driven image mechanics (D2IM): a deep learning approach to predict displacement and strain fields from undeformed X-ray tomography images - Evaluation of bone mechanics.

This submission contains the full dataset used for training and evaluating the D2IM model presented in the paper.

The 'Input' folder contains two subdirectories, one containing the clinical scans (input 1 in the model) and the binary mask made from these scans (input 2 in the model).
The 'Target' folder contains three subdirectories, which contain the ground truth images corresponding to the full field displacement data for the three displacement components u, v and w.
The 'Clinical' folder contains a few examples of clinical and resized images used (along with masks and other required imaging) used for testing the ability of the model to scale to lower resolution imaging.

Also contained in this submission is an example notebook file (d2im-prototype.ipynb) showing the code used to train and evaluate the model, as well as generate the statistics and images presented in the paper. This is also accompanied with two examples of trained models(D2IM_trained.h5, D2IM_trained_data_augmentation.h5) that can be used to make predictions from the images.