Supervised and unsupervised training using CIFAR-10 dataset
Dependencies:
python 2.7
keras '2.1.5'
matplotlib '2.2.2'
sklearn '0.20.0'
tensorflow '1.4.0'
train.py is the training program that trains autoencoder, initial classifier and final classifier; also saves all the model, their corresponding weights and figures to disk.
test.py is the evaluation program that evaluates the performance of the classifier.
How to Run:
python train.py --dataset cifar-10-batches-py --model_ae model_autoencoder --model_cl model_classifier --images images
python test.py --dataset cifar-10-batches-py --model_cl model_classifier --images images