def load_checkpoint(name):
check = torch.load("../models/" + model, map_location="cuda")
loss = check["val_best_loss"]
model = se_resnet34(num_classes=2, multi_output=True).to(device)
model = densenet121(if_selayer=True).to(device)
model.load_state_dict(check["model"])
model = model.to(device)
model.eval()
return model
def main():
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
data_transform = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
# load image
img_path = "../tulip.jpg"
assert os.path.exists(img_path), "file: '{}' dose not exist.".format(
img_path)
img = Image.open(img_path)
plt.imshow(img)
# [N, C, H, W]
img = data_transform(img)
# expand batch dimension
img = torch.unsqueeze(img, dim=0)
# read class_indict
json_path = './class_indices.json'
assert os.path.exists(json_path), "file: '{}' dose not exist.".format(
json_path)
json_file = open(json_path, "r")
class_indict = json.load(json_file)
# create model
model = densenet121(num_classes=5).to(device)
# load model weights
model_weight_path = "./weights/model-3.pth"
model.load_state_dict(torch.load(model_weight_path, map_location=device))
model.eval()
with torch.no_grad():
# predict class
output = torch.squeeze(model(img.to(device))).cpu()
predict = torch.softmax(output, dim=0)
predict_cla = torch.argmax(predict).numpy()
print_res = "class: {} prob: {:.3}".format(
class_indict[str(predict_cla)], predict[predict_cla].numpy())
plt.title(print_res)
for i in range(len(predict)):
print("class: {:10} prob: {:.3}".format(class_indict[str(i)],
predict[i].numpy()))
plt.show()
def main():
logger, result_dir, _ = utils.config_backup_get_log(args, __file__)
device = utils.get_device()
utils.set_seed(args.seed, device) # set random seed
dataset = COVID19DataSet(root=args.datapath,
ctonly=args.ctonly) # load dataset
trainset, testset = split_dataset(dataset=dataset, logger=logger)
if args.model.lower() in ['mobilenet']:
net = mobilenet_v2(task='classification',
moco=False,
ctonly=args.ctonly).to(device)
elif args.model.lower() in ['densenet']:
net = densenet121(task='classification',
moco=False,
ctonly=args.ctonly).to(device)
else:
raise Exception
criterion = torch.nn.BCEWithLogitsLoss()
optimizer = torch.optim.Adam(net.parameters(),
lr=args.lr,
weight_decay=1e-3)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer,
step_size=25,
gamma=0.1)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=args.bstrain,
shuffle=True,
num_workers=args.nworkers)
testloader = torch.utils.data.DataLoader(testset,
batch_size=args.bstest,
shuffle=False,
num_workers=args.nworkers)
best_auroc = 0.
print('==> Start training ..')
start = time.time()
for epoch in range(args.maxepoch):
net = train(epoch, net, trainloader, criterion, optimizer, scheduler,
args.model, device)
scheduler.step()
if epoch % 5 == 0:
auroc, aupr, f1_score, accuracy = validate(net, testloader, device)
logger.write(
'Epoch:%3d | AUROC: %5.4f | AUPR: %5.4f | F1_Score: %5.4f | Accuracy: %5.4f\n'
% (epoch, auroc, aupr, f1_score, accuracy))
if auroc > best_auroc:
best_auroc = auroc
best_aupr = aupr
best_epoch = epoch
print("save checkpoint...")
torch.save(net.state_dict(),
'./%s/%s.pth' % (result_dir, args.model))
auroc, aupr, f1_score, accuracy = validate(net, testloader, device)
logger.write(
'Epoch:%3d | AUROC: %5.4f | AUPR: %5.4f | F1_Score: %5.4f | Accuracy: %5.4f\n'
% (epoch, auroc, aupr, f1_score, accuracy))
if args.batchout:
with open('temp_result.txt', 'w') as f:
f.write("%10.8f\n" % (best_auroc))
f.write("%10.8f\n" % (best_aupr))
f.write("%d" % (best_epoch))
end = time.time()
hours, rem = divmod(end - start, 3600)
minutes, seconds = divmod(rem, 60)
print("Elapsed Time: {:0>2}:{:0>2}:{:05.2f}".format(
int(hours), int(minutes), seconds))
logger.write("Elapsed Time: {:0>2}:{:0>2}:{:05.2f}\n".format(
int(hours), int(minutes), seconds))
return True
# model = se_resnet34(num_classes=2).to(device)
model = se_resnet34(num_classes=2, multi_output=True).to(device)
elif args.model == "bengali_resnet34":
model = model_bengali.se_resnet34(num_classes=2,
multi_output=True).to(device)
elif args.model == "bengali_resnext50":
model = model_bengali.se_resnext50_32x4d(num_classes=2,
multi_output=True).to(device)
elif args.model == "resnet152":
model = se_resnet152(num_classes=2, multi_output=True).to(device)
elif args.model == "resnext50":
model = se_resnext50_32x4d(num_classes=2, multi_output=True).to(device)
elif args.model == "resnext101":
model = se_resnext101_32x8d(num_classes=2, multi_output=True).to(device)
elif args.model == "densenet":
model = densenet121(if_selayer=True).to(device)
elif args.model == "inception_v3":
model = torchvision.models.inception_v3(pretrained=False,
num_classes=11 + 168 +
7).to(device)
elif args.model in efficientnets:
from efficientnet_pytorch import EfficientNet
efficientnet_name = args.model
print("efficientnet:", efficientnet_name)
model = EfficientNet.from_pretrained(args.model,
num_classes=11 + 168 + 7).to(device)
else:
raise ValueError()
# train_all = load_train_df()
def train(args):
batch_size = args.batch_size
epoch = args.epoch
network = args.network
opt = args.opt
train = unpickle(args.train_path)
test = unpickle(args.test_path)
train_data = train[b'data']
test_data = test[b'data']
x_train = train_data.reshape(train_data.shape[0], 3, 32, 32)
x_train = x_train.transpose(0, 2, 3, 1)
y_train = train[b'fine_labels']
x_test = test_data.reshape(test_data.shape[0], 3, 32, 32)
x_test = x_test.transpose(0, 2, 3, 1)
y_test = test[b'fine_labels']
x_train = norm_images(x_train)
x_test = norm_images(x_test)
print('-------------------------------')
print('--train/test len: ', len(train_data), len(test_data))
print('--x_train norm: ', compute_mean_var(x_train))
print('--x_test norm: ', compute_mean_var(x_test))
print('--batch_size: ', batch_size)
print('--epoch: ', epoch)
print('--network: ', network)
print('--opt: ', opt)
print('-------------------------------')
if not os.path.exists('./trans/tran.tfrecords'):
generate_tfrecord(x_train, y_train, './trans/', 'tran.tfrecords')
generate_tfrecord(x_test, y_test, './trans/', 'test.tfrecords')
dataset = tf.data.TFRecordDataset('./trans/tran.tfrecords')
dataset = dataset.map(parse_function)
dataset = dataset.shuffle(buffer_size=50000)
dataset = dataset.batch(batch_size)
iterator = dataset.make_initializable_iterator()
next_element = iterator.get_next()
x_input = tf.placeholder(tf.float32, [None, 32, 32, 3])
y_input = tf.placeholder(tf.int64, [None, ])
y_input_one_hot = tf.one_hot(y_input, 100)
lr = tf.placeholder(tf.float32, [])
if network == 'resnet50':
prob = resnet50(x_input, is_training=True, reuse=False, kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnet34':
prob = resnet34(x_input, is_training=True, reuse=False,
kernel_initializer=tf.contrib.layers.variance_scaling_initializer())
elif network == 'resnet18':
prob = resnet18(x_input, is_training=True, reuse=False,
kernel_initializer=tf.contrib.layers.variance_scaling_initializer())
elif network == 'seresnet50':
prob = se_resnet50(x_input, is_training=True, reuse=False, kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnet110':
prob = resnet110(x_input, is_training=True, reuse=False, kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnet110':
prob = se_resnet110(x_input, is_training=True, reuse=False, kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnet152':
prob = se_resnet152(x_input, is_training=True, reuse=False, kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnet152':
prob = resnet152(x_input, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnet_fixed':
prob = get_resnet(x_input, 152, trainable=True, w_init=tf.orthogonal_initializer())
elif network == 'densenet121':
prob = densenet121(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'densenet169':
prob = densenet169(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'densenet201':
prob = densenet201(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'densenet161':
prob = densenet161(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'densenet100bc':
prob = densenet100bc(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'densenet190bc':
prob = densenet190bc(x_input, reuse=False, is_training=True, kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnext50':
prob = resnext50(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnext110':
prob = resnext110(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
elif network == 'resnext152':
prob = resnext152(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnext50':
prob = se_resnext50(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnext110':
prob = se_resnext110(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
elif network == 'seresnext152':
prob = se_resnext152(x_input, reuse=False, is_training=True, cardinality=32,
kernel_initializer=tf.orthogonal_initializer())
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prob, labels=y_input_one_hot))
conv_var = [var for var in tf.trainable_variables() if 'conv' in var.name]
l2_loss = tf.add_n([tf.nn.l2_loss(var) for var in conv_var])
loss = l2_loss * 5e-4 + loss
if opt == 'adam':
opt = tf.train.AdamOptimizer(lr)
elif opt == 'momentum':
opt = tf.train.MomentumOptimizer(lr, 0.9)
elif opt == 'nesterov':
opt = tf.train.MomentumOptimizer(lr, 0.9, use_nesterov=True)
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.control_dependencies(update_ops):
train_op = opt.minimize(loss)
logit_softmax = tf.nn.softmax(prob)
acc = tf.reduce_mean(tf.cast(tf.equal(tf.argmax(logit_softmax, 1), y_input), tf.float32))
# -------------------------------Test-----------------------------------------
if not os.path.exists('./trans/tran.tfrecords'):
generate_tfrecord(x_test, y_test, './trans/', 'test.tfrecords')
dataset_test = tf.data.TFRecordDataset('./trans/test.tfrecords')
dataset_test = dataset_test.map(parse_test)
dataset_test = dataset_test.shuffle(buffer_size=10000)
dataset_test = dataset_test.batch(128)
iterator_test = dataset_test.make_initializable_iterator()
next_element_test = iterator_test.get_next()
if network == 'resnet50':
prob_test = resnet50(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'resnet18':
prob_test = resnet18(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'resnet34':
prob_test = resnet34(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'seresnet50':
prob_test = se_resnet50(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'resnet110':
prob_test = resnet110(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'seresnet110':
prob_test = se_resnet110(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'seresnet152':
prob_test = se_resnet152(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'resnet152':
prob_test = resnet152(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'seresnet_fixed':
prob_test = get_resnet(x_input, 152, type='se_ir', trainable=False, reuse=True)
elif network == 'densenet121':
prob_test = densenet121(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'densenet169':
prob_test = densenet169(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'densenet201':
prob_test = densenet201(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'densenet161':
prob_test = densenet161(x_input, is_training=False, reuse=True, kernel_initializer=None)
elif network == 'densenet100bc':
prob_test = densenet100bc(x_input, reuse=True, is_training=False, kernel_initializer=None)
elif network == 'densenet190bc':
prob_test = densenet190bc(x_input, reuse=True, is_training=False, kernel_initializer=None)
elif network == 'resnext50':
prob_test = resnext50(x_input, is_training=False, reuse=True, cardinality=32, kernel_initializer=None)
elif network == 'resnext110':
prob_test = resnext110(x_input, is_training=False, reuse=True, cardinality=32, kernel_initializer=None)
elif network == 'resnext152':
prob_test = resnext152(x_input, is_training=False, reuse=True, cardinality=32, kernel_initializer=None)
elif network == 'seresnext50':
prob_test = se_resnext50(x_input, reuse=True, is_training=False, cardinality=32, kernel_initializer=None)
elif network == 'seresnext110':
prob_test = se_resnext110(x_input, reuse=True, is_training=False, cardinality=32, kernel_initializer=None)
elif network == 'seresnext152':
prob_test = se_resnext152(x_input, reuse=True, is_training=False, cardinality=32, kernel_initializer=None)
logit_softmax_test = tf.nn.softmax(prob_test)
acc_test = tf.reduce_sum(tf.cast(tf.equal(tf.argmax(logit_softmax_test, 1), y_input), tf.float32))
# ----------------------------------------------------------------------------
saver = tf.train.Saver(max_to_keep=1, var_list=tf.global_variables())
config = tf.ConfigProto()
config.allow_soft_placement = True
config.gpu_options.allow_growth = True
now_lr = 0.001 # Warm Up
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
counter = 0
max_test_acc = -1
for i in range(epoch):
sess.run(iterator.initializer)
while True:
try:
batch_train, label_train = sess.run(next_element)
_, loss_val, acc_val, lr_val = sess.run([train_op, loss, acc, lr],
feed_dict={x_input: batch_train, y_input: label_train,
lr: now_lr})
counter += 1
if counter % 100 == 0:
print('counter: ', counter, 'loss_val', loss_val, 'acc: ', acc_val)
if counter % 1000 == 0:
print('start test ')
sess.run(iterator_test.initializer)
avg_acc = []
while True:
try:
batch_test, label_test = sess.run(next_element_test)
acc_test_val = sess.run(acc_test, feed_dict={x_input: batch_test, y_input: label_test})
avg_acc.append(acc_test_val)
except tf.errors.OutOfRangeError:
print('end test ', np.sum(avg_acc) / len(y_test))
now_test_acc = np.sum(avg_acc) / len(y_test)
if now_test_acc > max_test_acc:
print('***** Max test changed: ', now_test_acc)
max_test_acc = now_test_acc
filename = 'params/distinct/' + network + '_{}.ckpt'.format(counter)
# saver.save(sess, filename)
break
except tf.errors.OutOfRangeError:
print('end epoch %d/%d , lr: %f' % (i, epoch, lr_val))
now_lr = lr_schedule(i, args.epoch)
break
def test(args):
# train = unpickle('/data/ChuyuanXiong/up/cifar-100-python/train')
# train_data = train[b'data']
# x_train = train_data.reshape(train_data.shape[0], 3, 32, 32)
# x_train = x_train.transpose(0, 2, 3, 1)
test = unpickle(args.test_path)
test_data = test[b'data']
x_test = test_data.reshape(test_data.shape[0], 3, 32, 32)
x_test = x_test.transpose(0, 2, 3, 1)
y_test = test[b'fine_labels']
x_test = norm_images(x_test)
# x_test = norm_images_using_mean_var(x_test, *compute_mean_var(x_train))
network = args.network
ckpt = args.ckpt
x_input = tf.placeholder(tf.float32, [None, 32, 32, 3])
y_input = tf.placeholder(tf.int64, [None, ])
# -------------------------------Test-----------------------------------------
if not os.path.exists('./trans/test.tfrecords'):
generate_tfrecord(x_test, y_test, './trans/', 'test.tfrecords')
dataset_test = tf.data.TFRecordDataset('./trans/test.tfrecords')
dataset_test = dataset_test.map(parse_test)
dataset_test = dataset_test.shuffle(buffer_size=10000)
dataset_test = dataset_test.batch(128)
iterator_test = dataset_test.make_initializable_iterator()
next_element_test = iterator_test.get_next()
if network == 'resnet50':
prob_test = resnet50(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'resnet18':
prob_test = resnet18(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'resnet34':
prob_test = resnet34(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'seresnet50':
prob_test = se_resnet50(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'resnet110':
prob_test = resnet110(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'seresnet110':
prob_test = se_resnet110(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'seresnet152':
prob_test = se_resnet152(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'resnet152':
prob_test = resnet152(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'seresnet_fixed':
prob_test = get_resnet(x_input, 152, type='se_ir', trainable=False, reuse=True)
elif network == 'densenet121':
prob_test = densenet121(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'densenet169':
prob_test = densenet169(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'densenet201':
prob_test = densenet201(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'densenet161':
prob_test = densenet161(x_input, is_training=False, reuse=False, kernel_initializer=None)
elif network == 'densenet100bc':
prob_test = densenet100bc(x_input, reuse=False, is_training=False, kernel_initializer=None)
elif network == 'densenet190bc':
prob_test = densenet190bc(x_input, reuse=False, is_training=False, kernel_initializer=None)
elif network == 'resnext50':
prob_test = resnext50(x_input, is_training=False, reuse=False, cardinality=32, kernel_initializer=None)
elif network == 'resnext110':
prob_test = resnext110(x_input, is_training=False, reuse=False, cardinality=32, kernel_initializer=None)
elif network == 'resnext152':
prob_test = resnext152(x_input, is_training=False, reuse=False, cardinality=32, kernel_initializer=None)
elif network == 'seresnext50':
prob_test = se_resnext50(x_input, reuse=False, is_training=False, cardinality=32, kernel_initializer=None)
elif network == 'seresnext110':
prob_test = se_resnext110(x_input, reuse=False, is_training=False, cardinality=32, kernel_initializer=None)
elif network == 'seresnext152':
prob_test = se_resnext152(x_input, reuse=False, is_training=False, cardinality=32, kernel_initializer=None)
# prob_test = tf.layers.dense(prob_test, 100, reuse=True, name='before_softmax')
logit_softmax_test = tf.nn.softmax(prob_test)
acc_test = tf.reduce_sum(tf.cast(tf.equal(tf.argmax(logit_softmax_test, 1), y_input), tf.float32))
var_list = tf.trainable_variables()
g_list = tf.global_variables()
bn_moving_vars = [g for g in g_list if 'moving_mean' in g.name]
bn_moving_vars += [g for g in g_list if 'moving_variance' in g.name]
var_list += bn_moving_vars
saver = tf.train.Saver(var_list=var_list)
config = tf.ConfigProto()
config.allow_soft_placement = True
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
saver.restore(sess, ckpt)
sess.run(iterator_test.initializer)
avg_acc = []
while True:
try:
batch_test, label_test = sess.run(next_element_test)
acc_test_val = sess.run(acc_test, feed_dict={x_input: batch_test, y_input: label_test})
avg_acc.append(acc_test_val)
except tf.errors.OutOfRangeError:
print('end test ', np.sum(avg_acc) / len(y_test))
break
def main(args):
device = torch.device(args.device if torch.cuda.is_available() else "cpu")
print(args)
print(
'Start Tensorboard with "tensorboard --logdir=runs", view at http://localhost:6006/'
)
tb_writer = SummaryWriter()
if os.path.exists("./weights") is False:
os.makedirs("./weights")
train_images_path, train_images_label, val_images_path, val_images_label = read_split_data(
args.data_path)
data_transform = {
"train":
transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
"val":
transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
}
# 实例化训练数据集
train_dataset = MyDataSet(images_path=train_images_path,
images_class=train_images_label,
transform=data_transform["train"])
# 实例化验证数据集
val_dataset = MyDataSet(images_path=val_images_path,
images_class=val_images_label,
transform=data_transform["val"])
batch_size = args.batch_size
nw = min([os.cpu_count(), batch_size if batch_size > 1 else 0,
8]) # number of workers
print('Using {} dataloader workers every process'.format(nw))
train_loader = torch.utils.data.DataLoader(
train_dataset,
batch_size=batch_size,
shuffle=True,
pin_memory=True,
num_workers=nw,
collate_fn=train_dataset.collate_fn)
val_loader = torch.utils.data.DataLoader(val_dataset,
batch_size=batch_size,
shuffle=False,
pin_memory=True,
num_workers=nw,
collate_fn=val_dataset.collate_fn)
# 如果存在预训练权重则载入
model = densenet121(num_classes=args.num_classes).to(device)
if args.weights != "":
if os.path.exists(args.weights):
load_state_dict(model, args.weights)
else:
raise FileNotFoundError("not found weights file: {}".format(
args.weights))
# 是否冻结权重
if args.freeze_layers:
for name, para in model.named_parameters():
# 除最后的全连接层外，其他权重全部冻结
if "classifier" not in name:
para.requires_grad_(False)
pg = [p for p in model.parameters() if p.requires_grad]
optimizer = optim.SGD(pg,
lr=args.lr,
momentum=0.9,
weight_decay=1E-4,
nesterov=True)
# Scheduler https://arxiv.org/pdf/1812.01187.pdf
lf = lambda x: ((1 + math.cos(x * math.pi / args.epochs)) / 2) * (
1 - args.lrf) + args.lrf # cosine
scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
for epoch in range(args.epochs):
# train
mean_loss = train_one_epoch(model=model,
optimizer=optimizer,
data_loader=train_loader,
device=device,
epoch=epoch)
scheduler.step()
# validate
acc = evaluate(model=model, data_loader=val_loader, device=device)
print("[epoch {}] accuracy: {}".format(epoch, round(acc, 3)))
tags = ["loss", "accuracy", "learning_rate"]
tb_writer.add_scalar(tags[0], mean_loss, epoch)
tb_writer.add_scalar(tags[1], acc, epoch)
tb_writer.add_scalar(tags[2], optimizer.param_groups[0]["lr"], epoch)
torch.save(model.state_dict(), "./weights/model-{}.pth".format(epoch))
# model selection
if args.model == "resnet18":
print("SeNet18")
model = se_resnet18(2,
if_mixup=args.mixup,
if_shake_shake=args.shake,
first_conv_stride=args.first_stride,
first_pool=True).to(device)
elif args.model == "resnet152":
print("SeNet152")
model = se_resnet152(2, if_mixup=args.mixup,
if_shake_shake=args.shake).to(device)
elif args.model == "densenet121":
print("DenseNet121")
model = densenet121(if_mixup=args.mixup,
if_selayer=args.se,
first_conv_stride=args.first_stride,
first_pool=True).to(device)
elif args.model == "densenet169":
print("DenseNet169")
model = densenet169(if_mixup=args.mixup, if_selayer=args.se).to(device)
elif args.model == "densenet201":
print("DenseNet201")
model = densenet201(if_mixup=args.mixup, if_selayer=args.se).to(device)
elif args.model == "dpn92":
print("DPN92")
model = dpn92(num_classes=2, if_selayer=args.se).to(device)
elif args.model == "dpn98":
print("DPN98")
model = dpn98(num_classes=2, if_selayer=args.se).to(device)
elif args.model == "dpn131":
print("DPN131")
def train(X_train,
Y_train,
X_val,
Y_val,
unlabeled_data=None,
batch=16,
gpu_id=0,
epoch=100):
"""
train model from training and validation data.
Args :
X_train : Training data
Y_train : Labels of training data
X_val : Validation data
Y_val : Labels of validation data
unlabeled_data : Training data which don't have labels (This is for semisupervised learning.)
batch : batchsize
gpu_id : GPU id where model is trained at
"""
assert len(X_train) == len(
Y_train
), "training data and its labels must have the same length : {}, {}".format(
len(X_train), len(Y_train))
assert len(X_val) == len(
Y_val
), "validation data and its labels must have the same length : {}, {}".format(
len(X_val), len(Y_val))
if unlabeled_data is not None:
assert X_train.shape[1:] == X_val.shape[1:] == unlabeled_data.shape[
1:], "All data must have the same shape"
H, W = X_train[0].shape[:2]
print("img shape: {}*{}".format(H, W))
BATCH_SIZE = batch
EPOCH = epoch
gpu_name = "cuda:" + str(gpu_id)
device = torch.device(gpu_name if torch.cuda.is_available() else "cpu")
torch.backends.cudnn.benchmark = True
print("=============training config=============")
print("device:{}".format(device))
print("MODEL:", args.model)
print("save dir : ", model_save_dir)
print("batch size:{}".format(BATCH_SIZE))
if args.mixup:
print("Manifold-mixup:{}".format(args.mixup))
if args.shake:
print("Shake-shake regularization:{}".format(args.shake))
if args.vat:
print("VAT regularization:{}".format(args.vat))
print("========================================")
# Calculate the average of pixels per channel
# def avg_each_channel(path):
# img = Image.open(path)
# if not len(np.array(img)) == 3:
# img = img.convert("RGB")
# img = np.asarray(img)
# ch_mean = np.average(np.average(img, axis=0), axis=0)
# return ch_mean # (3, )
#
# ch_means_per_image = Parallel(n_jobs=-1, verbose=-1)([delayed(avg_each_channel)(path) for path in path_dev])
# ch_mean = np.average(ch_means_per_image, axis=0)
# print("channel mean:{}".format(ch_mean))
#
# def std_each_channel(path):
# channel_mean = [237.78516378, 237.16343756, 237.0501237]
#
# img = Image.open(path)
# if not len(np.array(img)) == 3:
# img = img.convert("RGB")
# img = np.asarray(img)
# R, G, B = img[:, :, 0], img[:, :, 1], img[:, :, 2]
# R_flat = R.flatten()
# G_flat = G.flatten()
# B_flat = B.flatten()
#
# R_diff = np.sum(np.square(R_flat - channel_mean[0]))
# G_diff = np.sum(np.square(G_flat - channel_mean[1]))
# B_diff = np.sum(np.square(B_flat - channel_mean[2]))
#
# return (R_diff, G_diff, B_diff)
# pixels_diff = Parallel(n_jobs=-1, verbose=-1)([delayed(std_each_channel)(path) for path in path_dev])
# R_all, G_all, B_all = 0, 0, 0
# for pixel_diff in pixels_diff:
# R_all += pixel_diff[0]
# G_all += pixel_diff[1]
# B_all += pixel_diff[2]
#
# R_std = np.sqrt(R_all/(H*W*len(path_dev)))
# G_std = np.sqrt(G_all/(H*W*len(path_dev)))
# B_std = np.sqrt(B_all/(H*W*len(path_dev)))
#
# ch_std = [R_std, G_std, B_std]
# print("channel std:", ch_std)
ch_mean = [237.78516378, 237.16343756, 237.0501237]
ch_std = [146.47616225347468, 146.6169214951974, 145.59586636818233]
c_train = Counter(Y_train)
c_val = Counter(Y_val)
c_test = Counter(Y_test)
print("train:{}, val:{}, test:{}".format(c_train, c_val, c_test))
print(
"train data length:{}, validation data length:{}, test data length:{}".
format(len(X_train), len(X_val), len(X_test)))
if args.semisupervised:
train = MeanTeacherDataset(
X_train,
Y_train,
transform=transforms.Compose([
#ColorJitter(brightness=0.0, contrast=0.4, hue=0.0),
RandomRotate(hard_rotate=False, angle=5),
Regularizer(ch_mean, ch_std=ch_std),
#Samplewise_Regularizer(),
ToTensor()
]))
else:
train = Dataset(
X_train,
Y_train,
transform=transforms.Compose([ #UnsharpMasker(radius=5.0),
#ColorJitter(brightness=0.0, contrast=0.4, hue=0.0),
#RandomScalor(scale_range=(301, 330), crop_size=H),
RandomRotate(hard_rotate=False, angle=5),
Regularizer(ch_mean, ch_std=ch_std),
#Samplewise_Regularizer(),
ToTensor()
]))
val = Dataset(
X_val,
Y_val,
transform=transforms.Compose([ #UnsharpMasker(radius=5.0),
Regularizer(ch_mean, ch_std=ch_std),
#Samplewise_Regularizer(),
ToTensor()
]))
dataset_sizes = {"train": train.__len__(), "val": val.__len__()}
print("dataset size:", dataset_sizes)
val_batch = BATCH_SIZE
if args.pseudo_label:
# Class ratio may be unbalance
class_sample_count = np.array(
[len(np.where(Y_train == t)[0]) for t in np.unique(Y_train)])
weight = 1. / class_sample_count
samples_weight = np.array([weight[t] for t in Y_train])
samples_weight = torch.from_numpy(samples_weight)
samples_weight = samples_weight.double()
sampler = torch.utils.data.sampler.WeightedRandomSampler(
samples_weight, len(samples_weight))
train_iter = torch.utils.data.DataLoader(train,
batch_size=BATCH_SIZE,
shuffle=False,
sampler=sampler)
else:
train_iter = torch.utils.data.DataLoader(train,
batch_size=BATCH_SIZE,
shuffle=True)
val_iter = torch.utils.data.DataLoader(val,
batch_size=val_batch,
shuffle=False)
# model selection
if args.model == "resnet18":
print("SeNet18")
model = se_resnet18(2,
if_mixup=args.mixup,
if_shake_shake=args.shake,
first_conv_stride=2,
first_pool=True).to(device)
elif args.model == "resnet50":
print("SeNet50")
model = se_resnet50(2,
if_mixup=args.mixup,
if_shake_shake=args.shake,
first_conv_stride=2,
first_pool=True).to(device)
elif args.model == "resnet101":
print("SeNet101")
model = se_resnet101(2,
if_mixup=args.mixup,
if_shake_shake=args.shake,
first_conv_stride=2,
first_pool=True).to(device)
elif args.model == "resnet152":
print("SeNet152")
model = se_resnet152(2, if_mixup=args.mixup,
if_shake_shake=args.shake).to(device)
elif args.model == "densenet121":
print("DenseNet121")
model = densenet121(if_mixup=args.mixup,
if_selayer=args.se,
first_conv_stride=2,
first_pool=True,
drop_rate=args.drop_rate).to(device)
elif args.model == "dpn92":
print("DPN92")
model = dpn92(num_classes=2,
if_selayer=args.se,
if_mixup=args.mixup,
first_conv_stride=2,
first_pool=True).to(device)
elif args.model == "dpn98":
print("DPN98")
model = dpn98(num_classes=2, if_selayer=args.se,
if_mixup=args.mixup).to(device)
else:
print("WRONG MODEL NAME")
input("---------Stop-----------")
if args.semisupervised:
""" Declare the teacher model """
def sigmoid_rampup(current, rampup_length):
""" Exponential rampup from https://arxiv.org/abs/1610.02242 """
if rampup_length == 0:
return 1.0
else:
current = np.clip(current, 0.0, rampup_length)
phase = 1.0 - current / rampup_length
return float(np.exp(-5.0 * phase * phase))
def get_current_consistency_weight(epoch):
# Consistency ramp-up from https://arxiv.org/abs/1610.02242
return args.consistency * sigmoid_rampup(
epoch, rampup_length=int(args.epoch / 2))
def update_teacher(student, teacher, alpha, global_step):
"""
update parameters of the teacher model.
Args :
student : Current model to train
teacher : Current teacher model
alpha : A parameter of models mixing weights
global_step : Global step of training
"""
alpha = min(1 - 1 / (global_step + 1), alpha)
for teacher_param, param in zip(teacher.parameters(),
student.parameters()):
teacher_param.data.mul_(alpha).add_(1 - alpha, param.data)
teacher_model = copy.deepcopy(model)
consistency_criterion = nn.MSELoss()
print("model preparing...")
loss_function = nn.CrossEntropyLoss(ignore_index=-1)
bce_loss = torch.nn.BCEWithLogitsLoss()
softmax = torch.nn.Softmax(dim=1)
#Set optimizer
init_learning_rate = args.learning_rate
optimizer = optim.SGD(model.parameters(),
lr=init_learning_rate,
momentum=0.9,
nesterov=True,
dampening=0,
weight_decay=0.0005)
# optimizer = torch.optim.Adam(model.parameters(), lr=init_learning_rate)
scheduler = torch.optim.lr_scheduler.MultiStepLR(
optimizer, milestones=[int(EPOCH * 0.5),
int(EPOCH * 0.75)], gamma=0.1)
lowest_loss = 1000000000000
train_losses, val_losses = [], []
train_accs, val_accs = [], []
PATH = ""
global_step = 0
start_epoch = 0
Logger = Train_Logger()
if args.resume:
# resume training from the latest checkpoint
checkpoint_names = os.listdir(os.path.join("model", args.directory))
# find a checkpoint having lowest loss
min_loss = 1000000000000
for name in checkpoint_names:
loss = float(name.split("_")[0][7:])
if loss < min_loss:
model_name = name
min_loss = loss
MODEL_PATH = os.path.join("/home/tanaka301052/tegaki/model",
args.directory, model_name)
# load model and optimizer from the checkpoint
checkpoint = torch.load(MODEL_PATH, map_location=device)
model.load_state_dict(checkpoint['model_state_dict'])
model.to(device)
optimizer.load_state_dict(checkpoint['optim_state_dict'])
start_epoch = checkpoint['epoch'] + 1
lowest_loss = checkpoint['lowest_loss']
global_step = checkpoint["global_step"]
print("==>" + "model loaded:", model_name)
print("current epoch:", start_epoch)
print("lowest loss:", lowest_loss)
print("resuming...")
for epoch in range(start_epoch, EPOCH, 1):
scheduler.step()
# Training Phase
model.train()
train_loss = 0
train_corrects = 0
loss = 0
for i, train_data in enumerate(tqdm(train_iter)):
global_step += 1
if args.semisupervised:
# Get inputs for both student model and teacher model
samples1, samples2 = train_data
student_inputs, labels = samples1
teacher_inputs, labels = samples2
student_inputs = student_inputs.to(device)
teacher_inputs = teacher_inputs.to(device)
labels = labels.to(device)
# forwarding student
student_outputs = model(student_inputs)
_, preds = torch.max(student_outputs, 1)
# forwarding teacher
teacher_outputs = teacher_model(teacher_inputs).detach()
# classification loss for student
classification_loss = loss_function(student_outputs, labels)
# consistency loss between student and teacher
consistency_loss = consistency_criterion(
softmax(student_outputs), softmax(teacher_outputs))
# get weight of consistency loss
consistency_weight = get_current_consistency_weight(epoch)
# total loss
loss = classification_loss + consistency_weight * consistency_loss
optimizer.zero_grad()
loss.backward()
optimizer.step()
if epoch < int(args.epoch / 2):
# during ramp up.
alpha = 0.99
else:
alpha = 0.999
update_teacher(model,
teacher_model,
alpha=alpha,
global_step=global_step)
train_loss += classification_loss.item() * student_inputs.size(
0)
train_corrects += (preds == labels).sum().item()
else:
inputs, labels = train_data
inputs = inputs.to(device)
labels = labels.to(device)
if not args.mixup:
outputs = model(inputs)
loss = loss_function(outputs, labels)
_, preds = torch.max(outputs, 1)
train_corrects += (preds == labels).sum().item()
elif args.mixup:
lam = sample_lambda_from_beta_distribution(
alpha=args.mixup_alpha)
lam = torch.from_numpy(np.array(
[lam]).astype('float32')).to(device)
output, reweighted_target = model(inputs,
lam=lam,
target=labels,
device=device)
loss = bce_loss(output, reweighted_target)
train_loss += loss.item() * inputs.size(0)
optimizer.zero_grad()
loss.backward()
optimizer.step()
train_loss /= dataset_sizes["train"]
train_losses.append(train_loss)
train_acc = train_corrects / dataset_sizes["train"]
train_accs.append(train_acc)
print("=====> train loss:{:5f} Acc:{:5f}".format(
train_loss, train_acc))
# Validation Phase
model.eval()
val_loss = 0
loss = 0
val_corrects = 0
for i, val_data in enumerate(val_iter):
inputs, labels = val_data
inputs = inputs.to(device)
labels = labels.to(device)
outputs = model(inputs)
max_index = outputs.max(dim=1)[1]
loss = loss_function(outputs, labels)
val_corrects += (max_index == labels).sum().item()
val_loss += loss.item() * inputs.size(0)
val_loss /= dataset_sizes["val"]
val_losses.append(val_loss)
result = {
"epoch": epoch,
"train_loss": train_loss,
"val_loss": val_loss,
"lowest_loss": lowest_loss,
"val_accuracy": val_corrects / dataset_sizes["val"],
}
val_accs.append(result["val_accuracy"])
if val_loss < lowest_loss:
lowest_loss = val_loss
if epoch >= 0:
# Prevent saving the model in the first epoch.
# remove latest saved model
if os.path.isfile(PATH):
os.remove(PATH)
model_name = "valloss{:.5f}_epoch{}.model".format(
val_loss, epoch)
if not os.path.exists(model_save_dir):
os.makedirs(model_save_dir)
PATH = os.path.join(model_save_dir, model_name)
checkpoint = {
"epoch": epoch,
"global_step": global_step,
"model_state_dict": model.state_dict(),
"optim_state_dict": optimizer.state_dict(),
"lowest_loss": val_loss
}
torch.save(checkpoint, PATH)
line(
"Project:{} \nEPOCH:{} train loss:{:4f}, val Accuracy:{:4f}, val loss:{:4f}, best loss:{:4f}"
.format(args.directory, epoch, train_loss,
result["val_accuracy"], result["val_loss"],
lowest_loss))
print("EPOCH:{} val Accuracy:{:4f}, val loss:{:4f}, best loss:{:4f}".
format(epoch, result["val_accuracy"], result["val_loss"],
lowest_loss))
Logger.send_loss_img(str(args.directory),
train_losses,
val_losses,
process_name=args.directory)
Logger.send_accuracy_img(str(args.directory),
train_accs,
val_accs,
process_name=args.directory)
output_eval_file = os.path.join(
"result", "{}_training_log.txt".format(args.directory))
Logger.write_log(result, output_eval_file)
Logger.save_history(train_losses,
"save/{}_train_losses.pkl".format(args.directory))
Logger.save_history(val_losses,
"save/{}_val_losses.pkl".format(args.directory))
img_resized_array = np.array(img_resized, dtype="uint8")
return (img_resized_array, image_name)
image_information = Parallel(n_jobs=-1, verbose=-1)([
delayed(open_image)(name, args.images_path, args.image_size)
for name in image_names
])
image_holder, imagename_holder = [], []
for (image_array, image_name) in image_information:
image_holder.append(image_array)
imagename_holder.append(image_name)
model = densenet121(if_mixup=False,
if_selayer=True,
first_conv_stride=2,
first_pool=True).to(device)
checkpoint = torch.load(args.model_path, map_location=device)
model.load_state_dict(checkpoint['model_state_dict'])
model.to(device)
print("==>" + "model loaded:", args.model_path)
ch_mean = [237.78516378, 237.16343756, 237.0501237]
ch_std = [146.47616225347468, 146.6169214951974, 145.59586636818233]
# dummy labels
labels = [0] * len(image_holder)
test = Dataset(image_holder,
labels,
def train(epochs):
"""
训练模型
:param epochs:
:return:
"""
vgg = vgg16((None, 224, 224, 3), 102)
resnet = resnet50((None, 224, 224, 3), 102)
densenet = densenet121((None, 224, 224, 3), 102)
models = [vgg, resnet, densenet]
train_db, valid_db = load_db(32)
his = []
for model in models:
variables = model.trainable_variables
optimizers = tf.keras.optimizers.Adam(1e-4)
for epoch in range(epochs):
# training
total_num = 0
total_correct = 0
training_loss = 0
for step, (x, y) in enumerate(train_db):
with tf.GradientTape() as tape:
# train
out = model(x)
loss = tf.losses.categorical_crossentropy(
y, out, from_logits=False)
loss = tf.reduce_mean(loss)
training_loss += loss
grads = tape.gradient(loss, variables)
optimizers.apply_gradients(zip(grads, variables))
# training accuracy
y_pred = tf.cast(tf.argmax(out, axis=1), dtype=tf.int32)
y_true = tf.cast(tf.argmax(y, axis=1), dtype=tf.int32)
correct = tf.reduce_sum(
tf.cast(tf.equal(y_pred, y_true), dtype=tf.int32))
total_num += x.shape[0]
total_correct += int(correct)
if step % 100 == 0:
print("loss is {}".format(loss))
training_accuracy = total_correct / total_num
# validation
total_num = 0
total_correct = 0
for (x, y) in valid_db:
out = model(x)
y_pred = tf.argmax(out, axis=1)
y_pred = tf.cast(y_pred, dtype=tf.int32)
y_true = tf.argmax(y, axis=1)
y_true = tf.cast(y_true, dtype=tf.int32)
correct = tf.cast(tf.equal(y_pred, y_true), dtype=tf.int32)
correct = tf.reduce_sum(correct)
total_num += x.shape[0]
total_correct += int(correct)
validation_accuracy = total_correct / total_num
print(
"epoch:{}, training loss:{:.4f}, training accuracy:{:.4f}, validation accuracy:{:.4f}"
.format(epoch, training_loss, training_accuracy,
validation_accuracy))
his.append({
'accuracy': training_accuracy,
'val_accuracy': validation_accuracy
})
return his