To run, first install the dependencies in requirements.txt.

Make sure your dataset is in the following layout: /dataset root/[CLINIC NAME]/[EMBRYO ID]/[FOCAL PLANE].jpg.

Next, train the autoencoder for the perceptual loss by running python train_perceptual_ae.py. Beware: all paths have been replaced with PLACEHOLDER - you will need to update them first. Also, we use Weights and Biases to monitor our experiments. Be sure to keep an eye on them.

Once the autoencoder has been trained, you can train the generator networks by running python train_gan.py. Again, be sure to replace all the placeholders. The datasets folder contains PyTorch datasets for intermediate plane prediction (t4missingplanes.py) and upper/lower plane prediction (t4extrapolation.py) which should be exchanged in train_gan.py as appropriate.

Once you have your trained models (or for validation purposes while you're training models), you can make predictions using predict.py. Code for experiments can be seen in miccai.ipynb.