class_filename = filename_root + "_class.txt"
with open(os.path.join(class_path, class_filename), 'w') as f:
f.write(result_class_name)
# Load model
model = tf.keras.models.load_model('bacteria_model.h5', compile=False)
losses = {"class_output": "categorical_crossentropy", "segm_output": wbce}
lossWeights = {"class_output": 1.0, "segm_output": 1.0}
model.compile(optimizer='adam',
loss=losses,
loss_weights=lossWeights,
metrics=['accuracy'])
# Get training data
train_data = prepare_train_data(IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS, False)
X_train = train_data['X_train']
# Get testing data
test_data = prepare_test_data(IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS)
X_test = test_data['X_test']
# Get results
[class_predictions_train, segm_predictions_train] = model.predict(X_train,
verbose=1)
[class_predictions_test, segm_predictions_test] = model.predict(X_test,
verbose=1)
save_results_to_files(class_predictions_train, segm_predictions_train, 'train')
save_results_to_files(class_predictions_test, segm_predictions_test, 'test')
help='initial learning rate')
parser.add_argument('--epochs', type=int, default=10,
help='training epochs')
parser.add_argument('--seed', type=int, default=19,
help='random seed')
#for debug
parser.add_argument('--datanum', type=int, default=-1,
help='training only first N samples. (for debug)')
args = parser.parse_args()
## example of training from files & output model files
from prepare_data import get_sentences_from_files, prepare_train_data
sentences = get_sentences_from_files(args.datafiles)
train_data, train_target, word_to_idx, target_to_idx = prepare_train_data(sentences)
np.save(args.word_to_idx, word_to_idx)
np.save(args.target_to_idx, target_to_idx)
if args.datanum > 0:
train_data = train_data[:args.datanum]
train_target = train_target[:args.datanum]
trained_model = train_model(train_data, train_target, word_to_idx, target_to_idx,
model_type = args.type, embedding_dim = args.emdim, hidden_dim = args.hiddim, epochs = args.epochs, learning_rate = args.lr, seed = args.seed)
torch.save(trained_model, args.model)
parser.add_argument('--model_path', type=str, default='./logs/model_zoo/', help='model path')
parser.add_argument('--model_name_train', type=str, default="guo.h5", help='trained model name')
parser.add_argument('--model_name_predict', type=str, default="guo.h5", help='model used to predict')
parser.add_argument('--result_path', type=str, default="./result", help='model path')
parser.add_argument('--result_stats_path', type=str, default="./logs/statistic/", help='trained model name')
parser.add_argument('-t','--train_mode', type=lambda x: (str(x).lower() == 'true'), default=True, help='train the model or not')
parser.add_argument('-i','--nEpochs', type=int, default=2, help='number of epochs to train for')
parser.add_argument('-u','--upscale_factor', type=int, default=2, help="super resolution upscale factor")
opt = parser.parse_args()
if opt.train_mode:
print('===> Loading datasets')
train_data, train_label = prepare_train_data(opt.train_data_path, opt.upscale_factor)
print(train_data.shape)
print(train_label.shape)
test_data, test_label = prepare_test_data(opt.test_data_path, opt.upscale_factor)
print(test_data.shape)
print(test_label.shape)
data_all = [train_data, train_label, test_data, test_label]
print('===> Building model')
train(data_all, os.path.join(opt.model_path, opt.model_name_train), opt.nEpochs)
model_name_predict = opt.model_name_train
print('===> Testing')
stats = predict(os.path.join(opt.model_path, model_name_predict), opt.test_data_path, opt.result_path)
else:
print('===> Testing')
stats = predict(os.path.join(opt.model_path, opt.model_name_predict), opt.test_data_path, opt.result_path)
result_stats_save(stats, opt.result_stats_path)
import numpy as np
import scipy.io as sio
import gradient as gr
import prepare_data as pre
import sigmoid
import random
import math
import time
import csv
import datetime
import multilayer
train_small = sio.loadmat('train_small.mat')
train = sio.loadmat('train.mat')
test = sio.loadmat('test.mat')
train_small_data = train_small['train']
train_data = train['train']
test_data = test['test']
(features, labels) = pre.prepare_train_data(train_small_data)
(x_test, y_test) = pre.prepare_test_data(test_data)
if __name__ == '__main__':
ann = multilayer.NeuralNet()
(weights_ret, bias_ret) = ann.train_multilayer_SGD(labels[6], features[6], .01, 500, y_test, x_test)
import numpy as np
import scipy.io as sio
import gradient as gr
import prepare_data as pre
import sigmoid
import random
import math
import time
import csv
import datetime
import multilayer
train_small = sio.loadmat('train_small.mat')
train = sio.loadmat('train.mat')
test = sio.loadmat('test.mat')
train_small_data = train_small['train']
train_data = train['train']
test_data = test['test']
(features, labels) = pre.prepare_train_data(train_small_data)
(x_test, y_test) = pre.prepare_test_data(test_data)
if __name__ == '__main__':
ann = multilayer.NeuralNet()
(weights_ret, bias_ret) = ann.train_multilayer_SGD(labels[6], features[6],
.01, 500, y_test,
x_test)