#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import context
from src.models import autoencoder
from src.networks.autoencoder import AutoEncoder
if __name__ == '__main__':
# creation and training of an auto-encoder, to pre-process the data
ae1 = AutoEncoder(model=autoencoder.Model1((400, 200, 1)),
batch_size=4,
dataset_path='.') # Adapt to your path
ae1.compile()
hae1 = ae1.fit(epochs=10,
repeat=1,
fname='autoencoder-Model1',
fname_enc='encoder-Model1')
ae1.save_losses(hae1, 'encoder-Model1') # saving the losses
# Create the encoded dataset with the encoder
ae1.encode_dataset('.', '.')
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
######################################################################
# Train a new Auto-Encoder and save the encoder #
# #
######################################################################
import context
import os
from src.models import autoencoder
from src.networks.autoencoder import AutoEncoder
if __name__ == '__main__':
dataset_path = '../GREGOIRE/dataset_new' # '.'
name = 'Model1_test17'
# creation and training of an auto-encoder, to pre-process the data
ae1 = AutoEncoder(model=autoencoder.Model1((520, 480, 1)), batch_size=64, dataset_path=dataset_path) # Adapt to your path
ae1.compile()
hae1 = ae1.fit(epochs=2, repeat=1, fname=('autoencoder-'+name), fname_enc=('encoder-'+name))
ae1.save_losses(hae1, 'encoder-'+name) # saving the losses
# Create the encoded dataset with the encoder
#ae1 = AutoEncoder(load_models='encoder-Model1_test5', version=0)
if not os.path.exists(dataset_path + '/train_encoded_TS/'):
os.mkdir(dataset_path + '/train_encoded_TS/')
ae1.encode_dataset(dataset_path, dataset_path)