help=
'.csv file that stores the average and mean values required for normalization process'
)
parser.add_argument('-w',
'--workers',
default=1,
help='Number of workers for batch processing',
type=int)
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
test_stats = dnn.read_stats(args.dataloc + "test/stats")
test_dataset = dnn.DroneDataset(root_dir=args.dataloc + "test/",
transform=transforms.Compose([
dnn.RandomCrop(224),
dnn.Normalize(test_stats),
dnn.ToTensor()
]))
test_loader = DataLoader(test_dataset,
batch_size=args.batch,
shuffle=True,
num_workers=args.workers)
# Pretrained
network = models.resnet152(pretrained=True)
for param in network.parameters():
param.requires_grad = False
num_ftrs = network.fc.in_features
network.fc = nn.Sequential(nn.Linear(num_ftrs, 1), nn.Sigmoid())
network = network.to(device)
help=
'.csv file that stores the average and mean values required for normalization process'
)
parser.add_argument('-w',
'--workers',
default=1,
help='Number of workers for batch processing',
type=int)
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
val_stats = dnn.read_stats(args.dataloc + "val/stats")
val_dataset = dnn.DroneDataset(root_dir=args.dataloc + "val/",
transform=transforms.Compose([
dnn.Resize(224),
dnn.Normalize(val_stats),
dnn.ToTensor()
]))
val_loader = DataLoader(val_dataset,
batch_size=args.batch,
shuffle=True,
num_workers=args.workers)
network = models.resnet50(pretrained=True)
#for param in network.parameters():
# param.requires_grad = False
num_ftrs = network.fc.in_features
network.fc = nn.Sequential(nn.Linear(num_ftrs, 1), nn.Sigmoid())
network = torch.nn.DataParallel(network).to(device)
parser.add_argument('-o', '--open', default=None, help='Open the specified model for loading and further training')
parser.add_argument('-b', '--batch', default=5, help='Specified batch size for training', type=int)
parser.add_argument('-e', '--epochs', default=30, help='Number of epochs to train', type=int)
parser.add_argument('-l', '--lr', default=1e-5, help='Learning rate set for the optimizer', type=float)
parser.add_argument('-m', '--momentum', default=0.9, help='Momentum set for the optimizer', type=float)
parser.add_argument('-L', '--logdir', default="training_logs/", help="Folder where the training logs are saved")
parser.add_argument('-r', '--saverate', default=5, help='The interval to save model states', type=int)
parser.add_argument('-w', '--workers', default=4, help='Number of workers for batch processing', type=int)
parser.add_argument('-n', '--notes', default="", help='Additional notes regarding the training')
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
training_stats = dnn.read_stats(args.dataloc + "train/stats")
val_stats = dnn.read_stats(args.dataloc + "val/stats")
train_dataset = dnn.DroneDataset(root_dir=args.dataloc + "train/", transform=transforms.Compose([ dnn.Resize(300), dnn.Corrupt(0.2), dnn.FlipHzt(0.1), dnn.ContrastBrightness(3.0, 100, 0.1), dnn.Rotate(50.0, 0.1), dnn.RandomCrop(224), dnn.Normalize(training_stats), dnn.ToTensor()]))
train_loader = DataLoader(train_dataset, batch_size=args.batch, shuffle=True, num_workers=args.workers)
val_dataset = dnn.DroneDataset(root_dir=args.dataloc + "val/", transform=transforms.Compose([ dnn.Resize(224), dnn.Normalize(val_stats), dnn.ToTensor()]))
val_loader = DataLoader(val_dataset, batch_size=args.batch, shuffle=True, num_workers=args.workers)
network = models.resnet50(pretrained=True)
#for param in network.parameters():
# param.requires_grad = False
num_ftrs = network.fc.in_features
network.fc = nn.Sequential(nn.Linear(num_ftrs, 1), nn.Sigmoid())
network = torch.nn.DataParallel(network).to(device)
#optimizer = torch.optim.Adam(network.parameters(), lr=args.lr)
parser.add_argument('-d', '--dataloc', default="datasets/", help='Path to the datasets')
parser.add_argument('-S', '--save_dir', default="saved_models/", help='Path to save trained models')
parser.add_argument('-o', '--open', default=None, help='Open the specified model for loading and further training')
parser.add_argument('-b', '--batch', default=5, help='Specified batch size for training', type=int)
parser.add_argument('-e', '--epochs', default=30, help='Number of epochs to train', type=int)
parser.add_argument('-l', '--lr', default=1e-4, help='Learning rate set for the optimizer', type=float)
parser.add_argument('-s', '--stats', default='stats', help='Path to file that stores the average and mean values required for normalization process')
parser.add_argument('-n', '--basename', default='net', help='Base name used to identify the particular training set')
parser.add_argument('-r', '--saverate', default=5, help='The interval to save model states', type=int)
parser.add_argument('-w', '--workers', default=4, help='Number of workers for batch processing', type=int)
args = parser.parse_args()
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
stats = dnn.read_stats(args.dataloc + args.stats)
dataset = dnn.DroneDataset(root_dir=args.dataloc, transform=transforms.Compose([ dnn.Scale(0.44444), dnn.Normalize(stats), dnn.ToTensor()]))
loader = DataLoader(dataset, batch_size=args.batch, shuffle=True, num_workers=args.workers)
network = model.Net().to(device)
optimizer = torch.optim.Adadelta(network.parameters(), lr=args.lr)
print("Model created")
model_name = args.open
if model_name is not None:
dnn.load_model(network, optimizer, model_name)
## Start Training
dnn.train(args, network, optimizer, loader)