dhadjia1 / neurogan Goto Github PK

Author: Darian Hadjiabadi
email: [email protected]

------------------------------------------------------


I'll try to make the readme quick - it's just to describe file hierarchy.

gan/ - Contains code for building architecture and training a general
adversarial network

utils/ - Contains code for preprocessing and visualization

database-swc-full/ - Contains all swc files downloaded from NeuroMorpho

database-imgs-full/ - Contains 64x64 images for pyramidal and granule cells
after preprocessing

output-large/ - Contains outputs from DCGAN.

models/ - Contains the 'best' DCGAN models for each class of neuron. 
Can be read in and used to generate dendritic morphologies.


preprocessing.ipynb - shows preprocessing steps
granule(pyramidal)-gan-development.ipynb - steps to build DCGAN model
granule(pyramidal)-quantitaive-assessmnet.ipynb - visualize and plots 
    network weights. Use kernel MMD to quantify model quality
generalization-check.ipynb - Use kernel MMD to access overfitting for
    both granule and pyramidal DCGAN models
granule(pyramidal)-qualitative-assessment.ipynb - creates shuffled datasets
    of real and fake images for users to disern real from fake
generate_morphologies.ipynb - Loads in built models and generates images.


Additional: database-imgs-small, database-swc-small, output-small, testing-platform.ipynb were all used for testing