search

guilherme-pombo / CounterSynth

Code for Equitable modelling of brain imaging by counterfactual augmentation with 3D deep generative models
7 stars 1 forks source link

Model weights? #1

Open zapaishchykova opened 3 months ago

zapaishchykova commented 3 months ago

Hi there! Thanks so much for making code open-source? I have a couple of questions:

Thanks!

guilherme-pombo commented 3 months ago

Hello, thank you for the interest in this project. Unfortunately, this was a project I did during my PhD and no longer have access to the model weights. I can add information for your other two questions this weekend, if that's of any use? That way you can reproduce the training with your own data, if you'd like.

Let me know and I'll set aside some time this weekend to address it.

zapaishchykova commented 3 months ago

Hi! Thanks so much for your quick reply. I don't want to take out any of your weekend time for this and I completely understand that you might not have access to the old data anymore. If you have some examples/code readily available, I would gladly take them, but otherwise, please don't bother. Thanks!

guilherme-pombo commented 3 months ago

Hello, the model_train.py should just work out of the box to train the model. If you are looking for data loading examples specific to Medical imaging, I recommend having look at the MONAI package. More specifically at some of their tutorials here:

https://github.com/Project-MONAI/tutorials/tree/main

The datasets.py file provided in this repo should also work simply by replacing the random tensors with the 3D volumes corresponding to your brain imaging (e.g. load in a NIFTI, numpy array, etc.). Unfortunately the brain imaging data used for my paper is not publicly shareable since it comes from UK Biobank. In order to get access to it, you have to apply through your institution (it's free). Hope this helps!