lotka / autofaces

A project investigating the combination of supervised and unsupervised neural network structures.

autofaces

general requirements:

python 2.7.6
tensorflow

pip requirements:

numpy
matplotlib
tqdm
sklearn
yaml
ruamel.yaml

For GPU support CUDA is required, follow the TensorFlow CUDA set up guide.

Config Files

src/config contains various configuration files for different set ups, the following examples use test.yaml

To Run

cd src/

Run an experiment:

python main.py --device=cpu --config=config/test.yaml

This will run the test set analysis, to run it agian:

python test_set_analysis.py path_to_results model model=final or early

To visualize the results run the viewResults.ipynb notebook in the notebooks folder, to compare multiple runs use compareResults.ipynb.

To run multiple experiments:

First edit relevant section in main.py under the line: if args.compare == True and args.compare != None: Then run:

python main.py --device=cpu --config=config/test.yaml --compare=True

or to run just one of the generated experiments:

python main.py --device=cpu --config=config/test.yaml --compare=True --batch=N