def main():
parser = argparse.ArgumentParser(description="train VoxResNet")
parser.add_argument(
"--iteration", "-i", default=100000, type=int,
help="number of iterations, default=100000")
parser.add_argument(
"--display_step", "-s", default=1000, type=int,
help="number of steps to display, default=1000")
parser.add_argument(
"--gpu", "-g", default=-1, type=int,
help="negative value indicates no gpu, default=-1")
parser.add_argument(
"--input_file", "-f", type=str, default="dataset_train.csv",
help="csv file containing filename of image and its label, default=dataset_train.csv")
parser.add_argument(
"--in_channels", type=int, default=2,
help="number of input image channels, default=2")
parser.add_argument(
"--n_classes", type=int, default=4,
help="number of output classes, default=4")
parser.add_argument(
"--n_batch", type=int, default=1,
help="batch size, default=1")
parser.add_argument(
"--shape", type=int, nargs='*', action="store",
default=[80, 80, 80],
help="shape of input for the network, default=[80, 80, 80]")
parser.add_argument(
'--out', '-o', default='vrn.npz', type=str,
help='trained model, default=vrn.npz')
parser.add_argument(
"--learning_rate", "-r", default=1e-3, type=float,
help="update rate, default=1e-3")
parser.add_argument(
"--weight_decay", "-w", default=0.0005, type=float,
help="coefficient of l2norm weight penalty, default=0.0005")
args = parser.parse_args()
print(args)
train_df = pd.read_csv(args.input_file)
vrn = VoxResNet(in_channels=args.in_channels, n_classes=args.n_classes)
if args.gpu >= 0:
chainer.cuda.get_device(args.gpu).use() # Make a specified GPU current
vrn.to_gpu()
xp = cuda.cupy if args.gpu >= 0 else np
optimizer = chainer.optimizers.Adam(alpha=args.learning_rate)
optimizer.use_cleargrads()
optimizer.setup(vrn)
optimizer.add_hook(chainer.optimizer.WeightDecay(args.weight_decay))
for i in range(args.iteration):
vrn.cleargrads()
scalar_img, label_img = load.sample(train_df, args.n_batch, args.shape)
assert np.max(label_img) < args.n_classes
assert label_img.shape == (args.n_batch,) + tuple(args.shape), label_img.shape
x_train = Variable(xp.asarray(scalar_img))
y_train = Variable(xp.asarray(label_img))
outputs = vrn(x_train, train=True)
loss = 0
for out in outputs:
loss += F.softmax_cross_entropy(out, y_train)
accuracy_c1 = F.accuracy(outputs[0], y_train)
accuracy = F.accuracy(outputs[-1], y_train)
loss.backward()
optimizer.update()
if i % args.display_step == 0:
print("step %5d, accuracy_c1 %.02f, accuracy %.02f, cost %g" % (i, accuracy_c1.data, accuracy.data, loss.data))
vrn.to_cpu()
chainer.serializers.save_npz(args.out, vrn)
def main():
parser = argparse.ArgumentParser(description="train VoxResNet")
parser.add_argument("--iteration",
"-i",
default=10000,
type=int,
help="number of iterations, default=10000")
parser.add_argument("--display_step",
"-s",
default=1000,
type=int,
help="number of steps to display, default=1000")
parser.add_argument("--gpu",
"-g",
default=-1,
type=int,
help="negative value indicates no gpu, default=-1")
parser.add_argument(
"--input_file",
"-f",
type=str,
default="dataset.json",
help="json file of traininig dataset, default=dataset.json")
parser.add_argument("--n_batch",
type=int,
default=1,
help="batch size, default=1")
parser.add_argument(
"--shape",
type=int,
nargs='*',
action="store",
default=[80, 80, 80],
help="shape of input for the network, default=[80, 80, 80]")
parser.add_argument('--out',
'-o',
default='vrn.npz',
type=str,
help='parameters of trained model, default=vrn.npz')
parser.add_argument("--learning_rate",
"-r",
default=1e-3,
type=float,
help="update rate, default=1e-3")
parser.add_argument(
"--weight_decay",
"-w",
default=0.0005,
type=float,
help="coefficient of l2norm weight penalty, default=0.0005")
args = parser.parse_args()
print(args)
with open(args.input_file) as f:
dataset = json.load(f)
train_df = pd.DataFrame(dataset["data"])
vrn = VoxResNet(dataset["in_channels"], dataset["n_classes"])
if args.gpu >= 0:
chainer.cuda.get_device(args.gpu).use()
vrn.to_gpu()
xp = chainer.cuda.cupy
else:
xp = np
optimizer = chainer.optimizers.Adam(alpha=args.learning_rate)
optimizer.use_cleargrads()
optimizer.setup(vrn)
optimizer.add_hook(chainer.optimizer.WeightDecay(args.weight_decay))
for i in range(args.iteration):
vrn.cleargrads()
image, label = load.sample(train_df, args.n_batch, args.shape)
x_train = xp.asarray(image)
y_train = xp.asarray(label)
logits = vrn(x_train, train=True)
loss = 0
for logit in logits:
loss += F.softmax_cross_entropy(logit, y_train)
loss.backward()
optimizer.update()
if i % args.display_step == 0:
accuracy = [
float(F.accuracy(logit, y_train).data) for logit in logits
]
print(
"step {0:5d}, acc_c1 {1[0]:.02f}, acc_c2 {1[1]:.02f}, acc_c3 {1[2]:.02f}, acc_c4 {1[3]:.02f}, acc {1[4]:.02f}"
.format(i, accuracy))
vrn.to_cpu()
chainer.serializers.save_npz(args.out, vrn)