from config import *
from models import Model
from util import *
from samples import *
from random import shuffle
alex = Model().build_alexnet()
alex.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
data = list(dataset.items())
shuffle(data)
wave_data, loaded_paths, loaded_labels = load_wavs(conf, data)
train_size = int(len(loaded_paths) * 0.98)
train_XX, test_XX = wave_data[:train_size], wave_data[train_size:]
train_X, test_X = sounds_to_mels(conf, train_XX), sounds_to_mels(conf, test_XX)
train_Y, test_Y = loaded_labels[:train_size], loaded_labels[train_size:]
alex.fit([train_X], [train_Y],
epochs=conf.epochs,
batch_size=conf.batch_size,
verbose=conf.verbose,
validation_data=([test_X], [test_Y]))
alex.save("model/alexnet.h5")
dataset = YOLODataset('data/train',
S,
B,
C,
label_dict=label_dict,
to_tensor=to_tensor)
dataloader = BaseDataLoader(dataset=dataset, batch_size=args.batch_size)
len_train = len(dataset) * 8 // 10
len_val = len(dataset) - len_train
train_dataloader, val_dataloader = dataloader.split([len_train, len_val])
net = YOLO(S, B, C)
if args.gpu:
net = net.cuda()
net = nn.DataParallel(net)
criterion = Loss(S, B, C)
metric = Loss(S, B, C)
optimizer = optim.SGD(filter(lambda p: p.requires_grad, net.parameters()),
lr=args.lr,
momentum=0.9)
scheduler = lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)
model = Model(net)
model.compile(optimizer, criterion, metric, scheduler, label_dict)
model.fit(train_dataloader=train_dataloader,
val_dataloader=val_dataloader,
epoch=args.epoch,
use_gpu=args.gpu)
