gendata = generatorData()
train_datas, train_labels = gendata.gen_Data(batch_size=60)
test_datas, test_labels = gendata.gen_Data(batch_size=40)
train_datas_shapes = np.shape(train_datas)
train_labels_s = np.reshape(train_labels, (int(len(train_labels) / 4), 4))
train_hot = to_categorical(train_labels_s)
input_shapes = np.shape(train_datas)[1:]
model = CRNN(input_shapes).CRNN_model
model.summary()
model.compile(loss={
'ctc': lambda train_hot, y_pred: y_pred
},
optimizer=RMSprop())
history = model.fit(
[train_datas, train_hot,
np.ones(1) * 15, np.ones(1) * 37],
np.ones(1),
batch_size=10,
epochs=100,
verbose=2)
print('Num. of images:', len(mj_synth.all_image_paths))
print('All Train {} / All Val {} / All Test {}'.format(
len(mj_synth.annotation_train), len(mj_synth.annotation_val),
len(mj_synth.annotation_test)))
X_train, y_train, X_val, y_val, X_test, y_test = mj_synth.random_choice(
random_choice_rate=0.005)
print('Train {} / Val {} / Test {}'.format(len(y_train), len(y_val),
len(y_test)))
train_ds, val_ds, test_ds = mj_synth.create_datasets(
X_train, y_train, X_val, y_val, X_test, y_val)
# Model definition
crnn = CRNN()
crnn.compile(mj_synth.max_label_len)
# Train the model
ckpt = ModelCheckpoint(filepath=args.save_model_path,
monitor='val_loss',
verbose=1,
save_best_only=True,
mode='auto')
callbacks_list = [ckpt]
crnn.training_model.fit(x=[*train_ds],
y=np.zeros(len(train_ds[0])),
batch_size=args.batch_size,
epochs=args.epochs,
validation_data=([*val_ds],
[np.zeros(len(val_ds[0]))]),
verbose=1,
def train_model(spectr_dir,
train_csv,
test_csv,
model_type,
use_cache=False,
resume=False):
print('Loading data... ')
if not use_cache:
x_train, y_train = load_data_from_spectrogram_dir(
spectr_dir, ['h', 'p'], train_csv)
x_test, y_test_ori = load_data_from_spectrogram_dir(
spectr_dir, ['h', 'p'], test_csv)
print('Saving data to cache...')
np.save('train_data.npy', x_train)
np.save('train_label.npy', y_train)
np.save('val_data.npy', x_test)
np.save('val_label.npy', y_test_ori)
else:
print('Loading data from cache...')
x_train = np.load('train_data.npy')
y_train = np.load('train_label.npy')
x_test = np.load('val_data.npy')
y_test_ori = np.load('val_label.npy')
print(y_train.shape)
label_count = np.bincount(y_train)
print('Train label count: ', label_count)
from sklearn.utils import class_weight
y_train = list(y_train)
class_weights = class_weight.compute_class_weight('balanced',
np.unique(y_train),
y_train)
print('sample weight per class', class_weights)
print('-' * 130)
print('Model train')
print('-' * 130)
input_shape = (rows, cols, channels)
from model import ResnetBuilder, CRNN, simple_CNN
if model_type == "resnet18":
model = ResnetBuilder.build_resnet_18(input_shape, num_classes)
elif model_type == "resnet34":
model = ResnetBuilder.build_resnet_18(input_shape, num_classes)
elif model_type == "CRNN":
model = CRNN(input_shape, num_classes)
else:
model = simple_CNN(input_shape, num_classes)
optimizer = Adadelta(0.1, rho=0.7)
model.compile(loss='categorical_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
x_train = x_train.reshape(x_train.shape[0], rows, cols, channels)
x_test = x_test.reshape(x_test.shape[0], rows, cols, channels)
y_train = to_categorical(y_train, num_classes)
y_test = to_categorical(y_test_ori, num_classes)
### Load weights
if resume:
model.load_weights('./model/' + model_name)
#
checkpoint = ModelCheckpoint('./model/' + model_name,
monitor='val_acc',
verbose=1,
save_best_only=True,
mode='max')
early_stop = EarlyStopping(monitor='val_acc', patience=10, mode='max')
callbacks_list = [checkpoint, early_stop]
print('epochs', epochs)
model.fit(x_train,
y_train,
batch_size=batch_size,
epochs=epochs,
validation_data=(x_test, y_test),
verbose=1,
shuffle=True,
callbacks=callbacks_list,
class_weight=class_weights)
print('Report')
y_prob = model.predict(x_test)
y_pred = y_prob.argmax(axis=-1)
print('test_y', np.bincount(y_test_ori))
print(classification_report(y_test_ori, y_pred, target_names=CLASSES))
model.save(model_name)