def __init__(self):
self.img_rows = 128
self.img_cols = 128
self.channels = 3
self.n_features = 128
self.n_classes = 31
self.img_shape = (self.img_rows, self.img_cols, self.channels)
self.data_loader = DataLoader(img_res=(self.img_rows, self.img_cols),
n_classes=self.n_classes)
optimizer = Adam(0.0002, 0.5)
self.D_R = build_discriminator(self.img_shape)
self.D_F = build_feature_discriminator(self.n_features)
self.D_R.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
self.D_F.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
self.Refiner = build_refiner(self.img_shape, self.channels)
self.Feature = build_encoder(self.img_shape, self.n_features)
self.Classifier = build_classifier(self.n_features, self.n_classes)
self.D_R.trainable = False
self.D_F.trainable = False
self.Classifier.compile(loss='categorical_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
self.Classifier.trainable = False
self.GAN_1 = Sequential()
self.GAN_1.add(self.Refiner)
self.GAN_1.add(self.D_R)
self.GAN_1.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
self.GAN_2 = Sequential()
self.GAN_2.add(self.Refiner)
self.GAN_2.add(self.Feature)
self.GAN_2.add(self.D_F)
self.GAN_2.compile(loss='binary_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
self.GAN_3 = Sequential()
self.GAN_3.add(self.Refiner)
self.GAN_3.add(self.Feature)
self.GAN_3.add(self.Classifier)
self.GAN_3.compile(loss='categorical_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
def __init__(self, n_vocab, max_seq_len, z_dim, c_dim, emb_dim,
pretrained_emb, freeze_embeddings, flow, flow_type, E_args,
G_args, C_args):
super(RNN_VAE, self).__init__()
self.MAX_SEQ_LEN = max_seq_len
self.n_vocab = n_vocab
self.z_dim = z_dim
self.c_dim = c_dim
self.device = torch.device('cuda')
"""
Word embeddings layer
"""
self.emb_dim = emb_dim
self.word_emb = nn.Embedding(n_vocab, self.emb_dim, PAD_IDX)
if pretrained_emb is not None:
assert self.emb_dim == pretrained_emb.size(
1), 'emb dim dont match with pretrained'
self.word_emb = nn.Embedding(n_vocab, self.emb_dim, PAD_IDX)
# Set pretrained embeddings
self.word_emb.weight.data.copy_(pretrained_emb)
if freeze_embeddings:
self.word_emb.weight.requires_grad = False
'''
Initialize all the modules
'''
self.encoder = build_encoder('gru',
emb_dim=self.emb_dim,
z_dim=z_dim,
**E_args)
self.decoder = build_decoder(embedding=self.word_emb,
emb_dim=self.emb_dim + z_dim + c_dim,
output_dim=n_vocab,
h_dim=z_dim + c_dim,
**G_args)
self.classifier = build_classifier('cnn', self.emb_dim, **C_args)
# Intiialize flow
self.use_flow = flow > 0
if self.use_flow:
self.flow_model = build_flow(flow_type, flow, z_dim)
from models.classifier import build_classifier, sample_audio
import os
dirname = r"../audio/samples"
FRAMES = 44100
samples = []
classifier = build_classifier(dirname, 10, "svc")
for f in os.listdir(dirname):
if f.endswith(".wav"):
fn = os.path.join(dirname, f)
sample = sample_audio(fn, 10000, 0)
p = classifier.predict_sample(sample)
print("{0}: {1}".format(f, p[0]))
val_gen = CompositeSequence(
thresholds_csv=classifier_config.THRESHOLDS_CSV,
batch_size=classifier_config.BATCH_SIZE,
shuffle_pre_training=False,
shuffle_on_epoch=False,
image_shape=classifier_config.INPUT_SHAPE,
train_val='val'
)
print(f'Training generator with {len(train_gen)} batches')
print(f'Validation generator with {len(val_gen)} batches')
print((len(train_gen)+len(val_gen))*classifier_config.BATCH_SIZE)
# Build model
model = build_classifier(classifier_config.AET_PATH, freeze_to_layer='concatenate_1')
optimizer = Adam(0.0001)
model.compile(
optimizer,
loss=CategoricalFocalLoss(alpha=0.25, gamma=4),
metrics=[mean_iou]
)
# Prepare callbacks
tb_cback = TensorBoard(log_dir=log_dir, batch_size=classifier_config.BATCH_SIZE)
model_ckpt = ModelCheckpoint(
filepath=f'./{weights_dir}/best_classifier_model.hdf5',
monitor='val_mean_iou',
save_best_only=True
)
sgdr_cback = SGDRScheduler(
from models.classifier import build_classifier, sample_audio
import os
import time
train_dir = r"../audio/train_samples"
test_dir = r"../audio/test_samples"
FRAMES = 44100
samples = []
classifier = build_classifier(train_dir, 10, )
tests = 0
for f in os.listdir(test_dir):
if f.endswith(".wav"):
fn = os.path.join(test_dir, f)
sample = sample_audio(fn, 10000, 0)
p = classifier.predict_sample(sample)
print("##################################")
tests += 1
print("Untrained sample {0}".format(str(tests)))
print("Processing '{0}'".format(f))
time.sleep(1)
print(" - Result: {0}".format(p[0]))
